Prediction of Covid-19 using Machine learning methods

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Abstract

 

Machine Learning methods are useful for various sectors. This has the ability to handle vast amount of data set to produce results. Machine learning provides more number of algorithms for prediction. Those are more effective to produce results in various problems. The clinical decision support systems are implemented using the methods of machine learning. All these methods are useful for predicting the disease in early stage. These methods also provide supports for making analysis on the dataset to produce the useful patterns that are simple to understand.  Covid-19 is one the infectious disease that is spreading all around the world. Each and every day, the infectious cases are increasing. Still research works are going on to diagnosis this disease in early stage. There is no medicine for this disease. So the prediction of this method is more important. All the countries are predicting the disease based on the sample collection form the patient. The result delivery takes more than 2 days. This research work focuses the execution of machine learning methods that are useful to analyze the death and recovery rate of patient who affected covid-19. The simple neural network and support vector neural network methods are used for this implementation. This research work produced accuracy, sensitivity and specificity using confusion matrix. The sentiment analysis is used to find the emotions of people on this covid-19.

Keywords-Machine learning, sentimental analysis, decision support system, accuracy, specificity

 

 

 

 

 

 

 

 

 

Chapter-1

Introduction

Health care is one of the essential sector which saves human life from various diseases and physical complexities. Covid-19 is the pandemic situation for the whole planet which becomes the big threat for human society. The first positive case of Covid-19 is reported in Wuhan, Chine in the year 2019. Then this virus has been gulping the entire world in a very short period of time. Various research works are started for the effective prediction of covid-19 using different methodologies. But still, this Covid-19 is harmfully spreading. According to the report of John Hopkins University, there are 4,563,458 covid-19 positive cases. All the countries are implementing various strategies to reduce the spreading speed of this Covid-19 such as travelling restrictions, social distancing, quarantines and lockdowns etc. There is no vaccine and medicine for this convid-19 virus. Many countries supported the researchers and motivated them to find the medicine for covid-19. Some countries are struggling with this covid-19.Because; the prediction methodologies are more expensive and complex. The used methodologies are taking more than 24 hrs. to produce the results. The effectiveness of the methods depends on the accuracy and time. So the effective methods are required to produce the accurate results in a short processing time. Each and every growth of the research work for covid-19 is essential for providing better prediction methods. Then only the existing methods can be avoided and the advanced methods will be used to produce the effective results.

Corona virus is the shortened name for Novel corona virus infections, which is a respiratory sickness that spreads from individual to individual(Bhat R., 2020) . Corona virus is the seventh individual from Corona affected family that taints people It was first revealed in December 2019 in china .Excess of 140,000+ individuals have passed up to 2020 April, in on universally from the Corona virus, while multiple millions contaminations have been affirmed in many nations, as per the World Health Organization, subsequently the Corona virus is presently proclaimed a pandemic. After watching the multiplication of the flare-up, a few nations were hit sooner than others, however as of now has less cases than the as of late hit nations(Boldog P., 2020;). As per the WHO’s head of crises, these sensational contrasts show that the conduct of governments because of this scourge matters, yet in particular, residents’ reaction as well. Indeed, while the quantity of individuals who are being treated for corona virus is expanding constantly, a few residents don’t know about the genuine danger of this flare-up which clarifies its immediately spread everywhere on over the world, anyway others, terrified and urgent, fell fast into the snare of this inauspicious and surprisingly more dreadful ending it all .

About 150+ days got over, ever since the first case of corona reported in China and the rate of people getting affected around the world is only increasing exponentially rather than showing a significant drop in number of cases reported on daily basis.Connections among people during developing pandemics presents an unmistakable general wellbeing training task. The transmission rate is moderately high. Essential examination has assessed that one individual who has it can spread it to somewhere in the range of 2 and 2.5 others. The infection seems to spread effectively among individuals, and keeps on being found after some time about how it spreads. Information has demonstrated that it spreads from one individual to individual. The infection spreads by respiratory beads delivered when somebody with the infection hacks or wheezes. These beads can be breathed in or land in the mouth of an individual close by. It can likewise spread on the off chance that a man contacts a tainted surface and, at that point contacts their mouth, nose or eyes, in spite of the fact that this isn’t viewed as a fundamental way it spreads. Studies found that the rate is getting higher; with one case spreading to somewhere in the range of 4.7 and 6.6 others in fast rate.Wellbeing shoppers must be educated about a looming wellbeing danger. Be that as it may, there might be troubles in giving precise data with respect to the flare-up in the starting stage(El Zowalaty M.E., 2020). This is for the most part identified with the serious extent of vulnerability about the specific course of spreading, contaminations treatments, & possibilities of recuperation in an episode. Every nation wants to get ready for the general wellbeing correspondence systems, media and network commitment staff for a potential case in their nation, just as for the proper reaction on the off chance that it occurs. The legislatures should arrange interchanges with other reaction associations also, remember the network for reaction activities. World health organization stands prepared to organize with accomplices to help nations in their correspondence and reaction to network commitment. To guarantee a people-focused reaction to virus, an extending gathering of worldwide reaction associations. These associations are dynamic at the worldwide, provincial and nation level to guarantee that influenced populaces have a voice and are essential for the reaction(Goyal K., 2020). Guaranteeing that worldwide proposals and correspondence are tried and adjusted to neighborhood settings will assist nations with overseeing the virus episode. People groups’ reaction to the report about a spreading infectious illness is probably going to prompt expanded uneasiness and intensification of danger recognitions. Online media systems considered to be data channels from which people in general could gain data relating sickness on the occasion of irresistible ailment. These stages additionally empower basic and speedy sharing of data with family, companions, and neighbors progressively. For instance, the Malaysian heath sector have been transferring presents related on this virus to teach the open since Jan  and their health director is additionally dynamic all alone online social media  to clear disarray and questions for the open. This is significant when conventional types of media can’t give applicable and ideal data to the general population. Web-based media presently fills in as a significant, prompt data source however while the focal point of most recent data has been on the part of online media during irresistible malady flare-ups, still the inquiry that ought to be uncovered, how utilization of online media will responsible for the open’s passionate or non cognitive reaction, influence impression of danger(Wang H., 2020).

Ever since from that time social media irrespective of the platform became a source of information about the novel this virus which can be delivering either truth or lies about certain facts. In the recent times, we can see the exponential development in the utilization of printed examination, characteristic language preparing and other computerized reasoning procedures in the field of exploration and development(Jin, et al., 2019). The exponential ascent in computerized reasoning techniques for literary investigation is trailed by the colossal increment in open dependence via web-based media stage, for the key data as opposed to relying upon convention means, for example, paper organizations. These days, individuals express their temperaments and conclusion via web-based media stage on assorted social phenomena’s (natural perils, governmental issues, social patterns and so on.).Thus, web-based media turning into an open correspondence stage and turning into a significant wellspring of information for conducting various types of research on human behavioral trends.During this pandemic, there were several measures taken  by the government of different countries to prevent the spread of Covid-19 cases, such as  country wide lockdown, social distancing and many more and it should be noted how people are responding to such incidents. It have been found, from recent studies that social distancing, lockdown and people losing their jobs has immensely affected the emotional stability of the people. The social media pandemic hadcomewell before the disease pandemic, which have resulted in diversified spectrum of emotions in humans(Skoric, et al., 2020, 11, 187). The trick partis that, the world have witnessed disease pandemic before, but it has never witnessed it duringa social media era. The effect it can cause during the social media era in such pandemic situation is yetto be analyzed. Thus, it makes very crucial to understand what public sentiments under the influence of Covid-19. To understand this we made a sentimental analysis through which people can understand about the realities of the ailments, the spreading proportion and consciousness of this maladies that can be shared by various individuals all around the globe. This assessment investigation will be useful in getting the client mindfulness and their activities against the corona(Chen, et al., 2020).

Sentimental analysis technique is an administered AI issue. There are various sorts of sentimental analysis or SA including assumption investigation, detect emotions, assessment examination base on aspects and multilingual sentiment investigation. In twofold assessment arrangement, the potential classes are + ve and -ve. feeling arrangement in fine grained, total five gatherings (extremely -ve, -ve, impartial, +ve, and positive)(Rocha & Lopes Cardoso, 2018). SA is one of the most well known errands in regular language handling, and there has been a great deal of examination and progress in illuminating this undertaking precisely. Profound neural systems are generally utilized in notion extremity order; be that as it may, it regularly requires tremendous quantities of preparing information, and the size of preparing information differs essentially among areas.  This was discovered that the best strategy that supports the understanding of frameworks with promising speculation capacities is a double approach by module. There aresome installing layers called BERT intended to prepare significant two directional depictions from messages which are not labeled by mutually molding on both left and right setting in all layers. It is pretrained from a huge unaided book corpus, for example, Wikipedia. Fifteen percent of the words are there in the information arrangement are conceal out which is one of the goals of the layer. At that point, a profound bidirectional transformer encoder is taken care of by the whole arrangements with the goal that the model figures out how to foresee the veiled words(Almatarneh & Gamallo, 2019).

Twitter information has been utilized broadly for printed and feelings examination. In another occasion, an investigation examining client input for a French Energy Company utilizing in excess of seventy thousand total tweets distributed longer than a year , utilized a Latent Dirichlet Allocation calculation to recover fascinating experiences about the vitality organization, covered up because of information volume, by recurrence based sifting procedures.(Kretinin, et al., 2018) Poisson and negative binomial models were utilized to investigate Tweet notoriety also. A similar report additionally assessed the connection between subjects utilizing seven uniqueness measures and found that, the Euclidean separations performed better in recognizing related themes valuable for client based intelligent methodology. Likewise, surviving examination applying time aware knowledge extraction strategy showed techniques to find important data from immense measures of data posted on Face book and Twitter. The investigation utilized subject based summing up of Twitter information to investigate substance of exploration premium. Essentially, they applied a system which utilizes less nitty gritty synopsis to deliver great quality data. Past examination has additionally explored the value of twitter information to survey character of clients, utilizing Disk (Dominance, Influence, Compliance and Unfaltering quality) evaluation procedures. Comparable exploration has been utilized in data frameworks utilizing printed investigation to create structures for ID of human characteristics, remembering strength for electronic correspondence. Circle evaluation is helpful for data recovery, content choice, item situating and mental evaluation of clients. So likewise, a mix of mental furthermore, etymological examination was utilized in past exploration to remove feelings from multilingual content posted on online media(Vijayan, et al., 2017).

The fast spread of virus also, diseases made a solid requirement for finding productive examination techniques for understanding the progression of data and the advancement of mass assessment in pandemic situations(S. Oh, 2020. ). While there are various activities dissecting medical care, safeguard, care and recuperation, financial and arrange information, there has been moderately little accentuation on the examination of total individual level and online media interchanges. Some papers as of late distinguished basic angles for virus administration and monetary recuperation situations. In their industry-situated report, they stressed information the executives, following and instructive dashboards as basic segments of dealing with a wide scope of virus situations. There has been an exponential development in the utilization of printed investigation, NLP and other computerized reasoning procedures in research and in the advancement of uses. Notwithstanding quick advances in natural language processing, issues encompassing the restrictions of these techniques in translating natural significance in text remain. Analysts at the famous institution MIT, they exhibited how even the most late natural language processing systems can miss the mark and in this manner stay “powerless against ill-disposed content”. It is, along these lines, imperative to comprehend inborn constraints of text grouping strategies and pertinent calculations. Moreover, it is essential to investigate whether different exploratory, elucidating and grouping procedures contain complimentary cooperative energies which will permit us to influence the “entire is more noteworthy than the aggregate of its parts” rule in our interest for man-made consciousness driven bits of knowledge age from human interchanges. Studies in electronic business sectors illustrated the viability of Artificial intelligence in demonstrating human conduct under complex educational conditions, featuring the part of the idea of data in influencing human conduct(S. Oh, 2020. ).

The ascent in accentuation on Artificial intelligence strategies for literary examination and natural language processing followed the colossal increment in open dependence via online media for data, as opposed to on the conventional news offices. Individuals express their assessments, dispositions, and exercises via online media about differing social wonders (for instance: wellbeing, characteristic risks, social elements, and social patterns) because of individual availability, arrange impacts, restricted expenses furthermore, simple access. Numerous organizations are utilizing online platform to advance the item a & administration end-clients. Clients exchange the encounters and audits, making a rich supply of data put away as text correspondingly. Subsequently, web-based media and open correspondence stages are turning out to be significant wellsprings of data for leading exploration, with regards to fast improvement of data and correspondence innovation. Scientists and professionals mine gigantic printed and unstructured datasets to create bits of knowledge about mass conduct, contemplations and feelings on a wide assortment of issues, for example, item audits, political conclusions and patterns, persuasive standards and financial exchange opinion .tweets were first ordered utilizing opinion examination, and afterward the movement of the dread supposition was examined, as it was the most prevailing feeling over the whole tweets information. Consequently, the discoveries uncover that the view of Twitter clients was generally sure or nonpartisan at whatever point they utilized any of these two hash tags while tweeting. Clients communicated least negative assumptions for both the hash tags. It demonstrates that there has been less concern and association on part of individuals about the illness. The central point in regards to the noticeable quality of positive assumptions over online media during this pandemic is that the greater part of the individuals, in spite of being focused and under lockdown, valued the endeavors of their particular governments and bleeding edge warriors like the wellbeing laborers, police faculty, and so forth. The clients seem confident about ideal endeavors and activity by their administrations and authorities, for example, complete lockdown, social removing, and rehearsing appropriate cleanliness estimates, for example, washing hands oftentimes and utilizing liquor-based sanitizers would overcome the pandemic soon. Different measures and limitations have been emphatically discovered paying little heed to the difficulties(Bhat R., 2020).

Despite the fact that there was cynicism, dread, sicken, and trouble about the lockdown, the positive notions stuck out. Indian people are certain that they needed to straighten the bend and were focused on it. Following notion stand apart was trust. Apparently, the Indians confided in their administration and were most likely sure that legislature would actualize the lockdown effectively and see that no residents would battle for fundamental things during the lockdown and the administration would make game plans for the equivalent. A few tweet communicated complete amazement relate to the choice however generally, it appeared as though individuals were anticipating that should contain the spread of the infection and this was conceivable just through social removing and by rehearsing cleanliness estimates like washing hands much of the time utilizing cleanser or liquor-based Hand wash(El Zowalaty M.E., 2020).

This investigation was educated by research articles from different controls and subsequently, in this area, we spread writing survey of printed examination, SA, tweets and characteristic language preparing, and Artificial intelligence techniques. Artificial intelligence and key organizing of data attributes are important to address advancing social issues in huge information(Almatarneh & Gamallo, 2019).

Surviving exploration has inspected phonetic difficulties and has exhibited the adequacy of ML techniques, for example, SVM (Support Vector Machine) in distinguishing outrageous slant. The focal point of this investigation is on showing how usually utilized ML techniques can be applied, and used to add to characterization of estimation by shifting Tweets attributes, and not the improvement of commitments to new ML hypothesis or calculations(Kowsari, et al., 2019). In contrast to direct relapse, which is basically utilized for assessing the likelihood of quantitative boundaries, grouping can be successfully utilized for assessing the likelihood of subjective boundaries for parallel or multi-class factors—that is the point at which the expectation variable of intrigue is paired, all out or ordinal in nature. There are numerous order techniques (classifiers) for subjective information; among the most notable are Naïve Bayes, strategic relapse, direct and KNN. The initial two are explained upon underneath with regards to literary investigation(Bhat R., 2020)

Problem Description

The rapid spread of Corona virus is named as Covid-19. This is a disease which is caused by the virus. The research works showed that Covid-19 has clinical characteristics like SARS-CoV.  The major symptoms of this disease are fever and cough. The main focus of detecting this disease is monitoring the fever level of a person.  The initial patients affected by this Covid-19, reportedly indicated with a market in Wuhan. Then this disease is announced as a pandemic that was declared by WHO. The corona viruses are single-stranded positive-sense RNA virus that is known to contain the huge viral genomes, up to around 32 kbp in length. After increments in the quantity of corona virus genome arrangements accessible there are number of research works intended to predict using effective methodologies.  According to the report of Hopkins University, there are 4,563,458 positive cases in Covid-19. This report is produced in 16th May 2020. All the countries are trying to provide better treatments for the patients. The improved identification methods are required to provide the early treatment for the people who are in starting stage. Most of the countries are announcing different conditions for the people to control the spreading speed of this virus. All these conditions and restrictions are trying to save the people by executing restrictions in travelling, conditions in social gathering, announcing the use of social distancing and providing lockdown announcements. This disease provides great impact on economic social life a human. This virus provides dangerous impacts for developing and under developing countries. The spreading ratio is high in the cities where the people density is high. This disease is a challenging one even for the developed countries. Some people do not have any symptoms in this disease. The researchers are trying to provide suitable method for identification and trying to give vaccines to avoid this virus. The proper treatment for this disease is not identified yet.  The socio economic crises are framed and reported by the United Nations. The impact of this disease will vary from one country to another. This can give increased poverty and inequalities at world level. The development paths in the long-term will be affected by the choices of countries. The Global Humanitarian Response Plan indicates the socio economic response for critical conditions. This covid-19 will affect more than 23 million of people who are affected by physical and psychological impairments. This can also affects lungs and chronic heart of a person. There are unclear symptoms in this disease prediction. There are some complications in this disease prediction. The still evolving clinical picture, death of patient follow-up and incomplete data on the number of people affected by covid-19 make is complex to predict the convid-19 infections. The information which is used in this present situation for prediction will not be sufficient for future prediction techniques. This disease will affect different age group people that cannot be controlled. The people who are having complications in lungs function can have a great chance of getting this disease. The treatment for this kind of people is also so risky. Lungs are the major parts which are affected by the covid-19 viruses.

New technologies like Artificial intelligence (AI), Internet of Things(IoT), Big Data and Machine Learning play an important role in the healthcare industry to fight against the new diseases. In this paper the role of AI technology is used to analyze the COVID-19 and other pandemics. AI is the upcoming tool for identifying the infection of corona virus in the early stage and also monitors the patient’s condition. The decision making and treatment consistency can be done using some AI based algorithms. It can also be used for monitoring the health condition of the infected patients. It helps to track the crisis on different angle like medical, molecular and epidemiological applications. It helps to develop a proper treatment methods and prevention strategies and also helps in the development of drug and vaccine.

Advancement in NLP (Natural Language Processing) technology and the pandemic makes use of AI tools to find out more relevant to the user and to extract a specific finding. The tools aim to identify the best solution form other disciplines. The tools are still under development condition. These tools can’t be used for the clinical research and decision making. The COVID-19 data set is inadvertently growing and it is useful for AI based analysis. NLP is used to extract the semantic feature which serves as an input to the long term memory and for the deep learning model. It is also used to predict the transmission law and develop a trend that will be useful to establish a prevention and control measures for the COVID-19 pandemics.

After analyzing the biological sequence a traditional machine learning or advance deep learning is used or combination of both the techniques can be used based on the problem being addressed. A tradition clustering method, hierarchical clustering and density –based spa tic clustering can be used to find the origin of the virus genomic sequences. Fuzzy logic systems can be used in the prediction of protein structures.

The application of AI is widely used in the daily live in various ways. Now the contribution of AI is dealing with Corona virus disease. Various methods that are used in the application to fight against the COVID-19 is being identified to outbreak and outline the crucial situation of the unprecedented battle of COVID-19. AI and its application in medical biology along with image processing, data analytics, text mining and Natural language processing, IoT are used in the identification and prevention of COVID-19.

To identify the confirmed cases a modified stacked auto encoder based on deep learning is used for forecasting the COVID-19 in real-time. The auto encoder model has 4 layers the input layer, first latent layer, second latent layer and the output layer with nodes 8,32,4 and 1. In this model 8 data points are provided as n input for the network the latent from the second latent layer are processed before providing it to the clustering algorithm and group the cases into provinces.

Radiology images like X-ray, CT scans are used as an high dimensional data and the processing of these high dimensional data is done by using deep learning methods where CNN-based models are most suitable. CNN helps to identify the visual patterns and the features of the captured image without prior knowledge. Transfer learning methods are also applied to customize the CNN model for diagnosis of COVID-19 which has a large medical image dataset.

For the detection of COVID-19  a framework is obtained by the use of Smartphone’s, sensors, cameras, microphones, temperature and inertial sensors. To identify the disease machine learning algorithms are employed for learning the disease symptoms based on the collected information. It helps to design a low-cost and quick approach to indentify or detect the corona virus. The data gathered from the temperature-fingerprint sensor can be used to predict the fever level. The information collected by onboard inertial sensors can be used in fatigue detection, neck posture monitoring and headache level prediction. Audio data obtained using the microphone can be used to identify the type of cough.

Symptoms of covid-19

The symptoms of covid-19 are classified into three different categories. Those are

  • Most common symptoms
  • Less common symptoms
  • Serious symptoms

Most common symptoms of this disease are

  • High temperature of body (high fever)
  • Dry cough
  • Tiredness

The Less common symptoms are

  • Body pain
  • Sore throat
  • Diarrhea
  • Head ache
  • Loss of smell and taste
  • Rashes on skin

Serious Symptoms are

  • Difficulties in breathing or shortness of breathing
  • Chest pain or pressure
  • Loss of movement or speech

The early symptoms are the signs for this corona infection to the human. The immediate attention will be given to these symptoms to make test for prediction. In case the result is positive, the strong medial are will be provided to the people. The person must be in quarantine while taking treatments. The severe symptoms are more dangerous for the human life. The main issue is the corona virus mainly affects the lungs. But more than 80% of people are affected by this disease without having any serious symptoms.

The World Health Organization (WHO) formed an International Working group which is used to provide the guidelines for key ethical issues and processes that are important for member state to address. The expert group provides various technical supports for this WHO which includes the advice for ethical questions related to the process for Covid-19. This world health organization affects number people around the world. The report of WHO indicates that there are some symptoms to identify this infection of Corona Virus. And the special treatment is required to cure this virus infection.  Aged persons those who are suffering from cardiovascular disease, diabetes, chronic respiratory disease, and cancer may have a chance of serious illness.

Some prevention activities are suggested by World Health Organization

  • Suggested to wash the hands with soap or water or clean the hands using sanitizers
  • Suggested to maintain social distancing
  • Suggested the take the test when they are having relative symptoms of Covid-19
  • Suggested to make self-quarantine when they feel not well
  • Suggested to stop smoking and other activities which affects the functions of lungs
  • Suggested to avoid unwanted travel and gatherings
  • Suggested to cover the mouth and nose when coughing and sneezing

National Ethics Committee

Many nations created official bodies to provide guidelines for covid-19 prevention and treatment activities.

This research work focuses the sentiment analysis of COVID-19 around the world. Social data on web contains some real life events such as COVID-19 and etc. Many people including government organizations can get the opinion of various people through this social media. Based on sentiment analysis the facts about the diseases, the spreading ratio and the awareness of this disease can be shared by different people all around the world. This sentiment analysis can be useful for getting the user awareness and their actions against COVID-19. Cloud computing is one of the powerful technology which can be used to store the social media data like twitter data, face book shares and etc. Analyzing these tweets and messages which are shared by various users can be used to find different emotions related to this disease.

Research Questions:

  • What is the aim of analyzing COVID-19 Messages?
  • Which methodology is used to collect the messages?
  • What is the expected outcome of this research?
  • What are all the supported methods for this analysis?
  • How the collected messages are handled for research?

 

Chapter-2

Literature Survey

Artificial Intelligence is the effective method which is highly used for pattern learning and prediction of COVID-19 (Narinder Singh Punn, 2020). The functions of artificial intelligence are used for describing the COVID-19 infections and providing the controls for social impacts. The deep learning methods are used for prediction of COVID-19 using pre-trained models. The self-learning methods are used for finding the future risks and improve the prediction of disease in early stage.

Classification using CT images are one of the popular methods of machine learning model. Using this famous method the Corona virus classification was done(Mucahid Barstugan., 2020). Various images related to covid-19 were collected and processed using different CT tools.  Five feature extraction methods were used for getting high accuracy in the prediction of Covid-19 such as Grey level Co-occurrence matrix, Local Directional Pattern, Grey level run length matrix, Grey level size zone matrix and Discrete Wavelet transform. Support vector classification method was used to classify the diseases using images. The proposed method achieved 99.68% of accuracy in the prediction of Covid-19.

The artificial neural network learning models are used for decision making in healthcare sectors (Mohammad Pourhomayoun., 2020). For this study, 117,000 records were collected from all over the world. Various machine learning model are used for analyzing the performance of prediction. Neural network, Random forest, decision tree, Support Vector Machine (SVM), Logistic regression and K-NN methods were used. Among the performances, the neural network achieved high performance than the other models. Confusion matrix was used for calculating the accuracy. 93% of accuracy was reached using neural network for the prediction of Covid-19.

Seven significant applications of AI for COVID-19 is identified by (Rajuj Vaishya, 2020) which plays an important role in the detection of cluster of cases and helps to predict what are the affect in future by collecting and analyzing all the previous data. The seven applications used are Early detection and diagnosis of the infection, Monitoring the treatment, Contract tracing of the individual, Projection of cases and mortality, Development of drugs and vaccines, Reducing the workload of healthcare works and Prevention of the disease

Social media is used for sentiment analysis for Covid-19 outbreak. Artificial Intelligence methods were used for analyzing the pattern prediction. For this analysis 92, 646 twitter messages were collected (Muzafar Bhat., 2020). Specifically 85,513 tweets were posted with hash tag Covid-19. Using sentiment analysis methods for 48,157 tweets, the positive sentiments are 51.97%, the neutral sentiments are   34.05% and the negative sentiments are 13.96%. They also specified the sentiment analysis for 35,296 tweets, 40.91% is for positive sentiment, and 17.80% is for negative sentiment. The remaining is for neutral.

The intelligent computing methods are used for Prognosis of Covid-19(Swapna rekha Hanumanthu., 2020). Various research methods were reviewed in this work. The machine learning models and deep learning models were taken for analysis. The performance results were compared with the accuracy, sensitivity and specificity produced by random forest, Support vector machine, K-NN, Logistic Regression, Convolution Neural Network(CNN) and Recurrent Neural Network (RNN). Among all these methods, the Support Vector machine model produced high accuracy in prediction of Covid-19. This model achieved 97.48% classification accuracy.

Machine learning methods were used for accurate prediction of Covid-19 using the blood test results(Jiangpeng Wu., 2020). 49 clinical available blood samples are used for prediction of Covid-19. These blood samples were extracted using random forest method. By using this method 95.95% of accuracy is reached. The Discriminative tool was used for execution of Covid-19 prediction. The proposed method also achieved 95.12% of sensitivity and 96.96% of specificity by using the blood sample data.

 

The author ThanhThi Nguyen states that artificial intelligence is used in wide range of areas for many problems. The artificial intelligence is used in image processing, natural language processing, IOT etc. It is used in wide range of areas. The artificial intelligence is applicable in finding solution for the pandemic covid-19. There are different methods in artificial intelligence and the author has suggested some methods which help the artificial researchers to find the solution for the covid-19 problem (Nguyen., 2020). The author has stated the different needs of data and different methods of artificial intelligence to handle the covid-19 problem.

The author has discussed about one of the artificial methods that are deep learning which helps to identify the problem accurately. The deep learning which contains many layers which helps to identify the image at higher level. It exactly works like human brain. The convolution neural network is another technique in deep learning which helps to identify the image. The convolution neural network contains many layers to identify the image. It contains different types of architecture. The author also states the different needs of data to find solution for the covid-19.

The author Vinay chamola et al states that the whole world is suffered due to the covid-19 that causes more deaths. Most of the countries were under lock down. Many of them were died due to this pandemic disease. So author has suggested some techniques to find the solution for covid-19 problem. The artificial intelligence and IOT were the two major technologies which help to recover from the covid-19 problem (Vinay Chamola., 2020). The author discussed about the different types of flu from the history and their impacts. The different types of industries that were affected by covid-19 were clearly discussed. The food industries, oil industries, agriculture and many industries were affected due to this covid-19. The symptoms of covid-19 were cough, fever, head ache etc.

The government of different nations was announced different rules and regulations to save people from the covid-19. The covid-19 has changed many lives and many people were suffered due to lack of food. Many prevention measures have taken to save the people. Different medicines were used to secure the people from covid-19 but till now there is no vaccination for the covid-19. The covid-19 is initially started its first step in china. Later it is spread among the world and whole world is suffering from the covid-19. Many 5G technologies, block chain technique, machine learning are good areas which help to recover from the covid-19 problem.

The author Nasir Saeed et al states that the use of wireless technologies has increased due to covid-19. The education systems are using the video conferencing and online classes for the students due to covid-19. The wireless connection has gained more importance during this pandemic era. The whole world is connected only by using the wireless. The wireless not only played the major role in education field but in the field of finding solutions for the covid-19 problem. The government imposed many rules and regulations to control the covid-19 but many of them lost their lives and home. The death rate is increased day by day and many industries are affected due to covid-19 (Nasir Saeed., 2020). This reflects the reduction in the economic growth and the government also struggles to save the people and to replace the normal condition. The wireless also helps in monitoring the spread of covid-19 and it also played an important role in health sectors. While talking about wireless the covid-19 is the biggest plus for the usage of wireless but there is also a biggest challenge in privacy and security in wireless. The person has to fight for this feature to secure the personal information. The IOT is the one of the important technology to find the affected person of covid-19 with the help of many sensors and tools. The emerging of 5G is the biggest challenges for many developers and they are keep on working to develop the 5G. To implement IOT it requires many power and uninterrupted power to automate the work. In many industries the automation is performed to overcome the problem of covid-19. There are many industries that are at risk and many industries will require automation due to covid-19. Thus the wireless technologies play a major role today and it will help to face the problem caused by covid-19. Thus it helps lot a people to recover from the flu.

The author RuchaVisal et al states that the social distancing is the one of the important preventive measure for the covid-19. The covid-19 is the very dangerous disease that has changed the life of many human beings (Rucha Visal ., 2020). The social distancing can be found by using the computer vision that is by using the camera and deep learning method. Where ever the crowd arises the system will give alert to the nearby police man or other authorities to clear the crowd. The CCTV camera will capture the image and the distance between two persons is calculated by using the deep learning method and the alert will be given. This will help to reduce the covid-19 spreading. The deep learning is the subset of the machine learning which works like the human brain and it will provide higher level of accuracy.

The convolution neural network is another technology which helps to find the covid-19 disease. X ray and other medical images are used to find the disease. The symptoms like cough, fever and body pain are the main symptoms for the covid-19. The government also introduced various rules to control the disease. The problem is not rectified till the government is facing the lot of problems and struggles due to covid-19. The vaccination is still in the third stage of its development. The mask is the important one to secure one from the covid-19 and many advertisements are given to make aware of the covid-19 spreading. The social distancing is the very important prevention to secure from the covid-19 disease.

The machine learning and artificial intelligence methods are used for identification of Covid-19(Lalmuanawma, 2020). This paper produced review about the machine learning methodologies and the artificial intelligence methods that are used for predicting the Covid-19 issues. They also discussed the challenges and risks in the covid-19 prediction. This paper also addresses the use of machine learning techniques and Artificial Intelligence techniques in the Covid-19 prediction. They highlighted the MYCIN expert system for the disease prediction using machine learning methods in healthcare industry. Fast diagnosis is possible by using this ML and AI methods for the pandemic disease.

Ahmad Alimadadi et al presented the machine learning models to fight Covid-19. They also highlighted that the report from Corona virus Resource Centre at Johns Hopkins university. This report showed the death ratio due to this Covid-19 (Alimadadi, 2020). The Artificial Intelligence Researchers involving to develop a prediction model using data mining techniques. The advanced machine learning methods are used for analyzing the virus spread and the diagnosis accuracy. The personalized protective strategies are used to categorize the people based on Covid-19 susceptibility. This research work involved in different data collections such as epidemiological data, clinical data and genetic data. After collecting the data, the data management process involved in the use of artificial intelligence methods, machine learning methods and deep learning method for Covid-19 prediction.

Parul Arora et al presented the paper that the deep learning models are used for covid-19 prediction and analysis. This case study was generated based on the covid-19 issues in India (Parul Arora., 2020). They reported that the deep learning methods are used for prediction of covid-19 cases in 32 states in India. The recurrent neural network model with long-short term memory was used for prediction of covid-19 in India. This paper also highlighted that deep LSTM and concurrent LSTM methods are applied on the Indian dataset for the prediction of Covid-19. The execution results showed that the proposed model produced reduced error in the daily and weekly prediction. The execution of the proposed model includes data collection, Transformation, training models, Hyper parameter tuning, Selection of models and future predictions. The error rate for daily prediction is less than 3% and the error rate for weekly prediction is less than 8%. So this provided the enhanced model for covid-19 prediction. This proposed method used forward pass and backward pass calculations to make speed analysis and accurate predictions. With the help of these methods, the spreading ratio is also calculated. The hot spot places were identified by using this analysis.

George Pinter et al proposed the hybrid machine learning approach which was used for covid-19 prediction in Hungary (Gergo Pinter., 2020). The advanced prediction methods are required to start the early treatment for the patients. The adaptive network based fuzzy inference system and multi-layer perception models are used for prediction and analysis process of covid-19 data that are collected from Hungary. The data was collected from the cases on March 4 to March 28. The spread ratio is calculated using machine learning methods and the accurate prediction is done by using the advanced concepts in machine learning. The proposed machine learning methods used MLP and ANFIS methods for getting better accuracy in the diagnosis of Covid-19. The implementation results showed that the machine learning methods are effective in analysis and prediction of covid-19. The existing methods are not efficient when compared with these proposed machine learning methods. In future, they will focus to provide better accuracy in prediction and effective results in analysis using convid-19 positive cases.

Amir Ahmed et al presented a use of machine learning methods for the prediction of Covid-19. They also discussed the methods and challenges in this prediction process (Amir Ahmad., 2020). Various machine learning approaches was used for identify the spreading level of this deadly disease. Susceptible Infectious Recovered model with its extension models are used for the prediction process. Those are Susceptible Infectious Recovered Deceased model and Susceptible Exposed Infectious Recovered model. By using these methods with differential equations are used for predicting the spreading covid-19 disease. In this case, the machine learning models considered the historical data for making the current prediction. Regression, clustering, classification and deep learning methods are used for different domains for providing improved results. The traditional machine learning regression and deep learning regression models are used for prediction of covid-19. This paper pointed that the main challenge of covid-19 prediction was insufficient data set. With the small amount of data, it was difficult to make more accurate prediction and provide training for machine learning methods.

Samir Kumar et al presented the machine learning approaches for positive cases of Covid-19. These methods are also used to predict the negative cases and the death rate of this disease (Samir Kumar Bandyopadhyay Sr., 2020). This research work proposed a verification method which is based on the deep learning techniques. For this research work, the Long Short Term Memory (LSTM) and Gated Recurrent Unit methods were used. These methods are effective to provide training for the machine learning models for improved prediction. The verification methods are used to make comparison between the original data and some pre-defined metrics. The proposed methods are effectively predicting the covid-19 cases. The execution results showed that the proposed models are efficient to produce the enhanced accuracy in covid-19 prediction using clinical data. These methods are useful for predicting the covid-19 cases with more effective manner. These methods are reducing the time taken for diagnosis and the accuracy produced by these methods are comparatively high than the existing methods.

ShreshthTuli et al proposed the use of Machine learning and cloud computing methods that are used for predicting the growth of covid-19 disease. This research work produced the improved mathematical for producing the better accuracy in prediction (ShreshthTuli., 2020).  The machine learning based improved model is used to predicting the positive cases and analyzing the current ratio of this disease in different countries. The improved model used iterative weighting for fitting Generalized Inverse Weibull distribution. This method is used to provide the better fit for developing a prediction model. For analysis, data are taken from different sources like text messages, online communications, social media messages and articles in web. By applying this machine learning methods on these data, the analysis report produced the spread of covid-19 and the hot spot places of this disease.  The CT image analysis using machine learning methods are helpful to monitor the patient’s health condition. For this research work, the dataset collected from Our World in data produced by Hannah Ritchie. This dataset is updates by the reports of world health organization. The execution results highlighted that the used machine learning models are helpful for the prediction and analysis of Covid-19.

Milind Yadav et al proposed the analysis process of covid-19 with the help of machine learning methods (MilindYadav., 2020). This research work aimed to produce various processes like prediction of covid-19 in different regions, growth rate analysis in various countries, research improvements for the prediction of covid-19 and analyzing the virus transmission rate. This research work also considers the weather condition for correlation of corona virus. This work analyze the minimizing the virus spreading speed by using various mitigation techniques. This analysis also focused the effectiveness of mitigation techniques with considering the working efficiency and protected case levels. For this research work, the support vector regression method is used. This is one of the supervised learning algorithm which is used for both classification and regression processes. This support vector regression is a non-parametric technique which is used to find the hyperplane in a dimensional space. This method is used for predicting the positive cases, deaths and recoveries of Covid-19.   With the use of this proposed method, the research work was used for predicting the spread of covid-19 in various regions. This also provides the different type of mitigation for this covid-19. The mitigation type includes lock down, closing schools; ban the large gathering and etc.

BirajaGhoshal et al resented the estimation of uncertainty and interoperability for deep learning methods to predict the covid-19 (Biraja Ghoshal, 2020). The effectiveness of deep learning methods is used for estimating the uncertainty and interoperability. These are used for computer based healthcare which used clinical data for disease diagnosis. The clinical dataset includes x-rays and other medical images which are used to produce more accurate predictions in the covid-19 cases. The Bayesian convolution neural network method was used for this research work. The deep learning methods are efficient to handle uncertainty and artificial intelligence methods are used for clinical setting. This proposed work is helpful for prediction, assessment and treatment for the positive cases. The chest x-ray dataset is used for covid-19 prediction process.

Xiang Bai et al proposed the prediction method for covid-19 malignant progression with Artificial Intelligence techniques (Xiang Bai., 2020). The CT scan dataset are considered for the prediction process. Various tests are applied on this dataset for effective prediction such as Chi-square test, Fisher’s exact test and T test. Traditional Logical Regression and Deep learning methods are used for prediction of covid-19. To test the performance of proposed methods the ROC curve was calculated. The experimental results showed that the deep learning methods are effective to produce the prediction results.

GurjithS.Randhawa et.al proposed a novel classification of Covid-19 using intrinsic genomic signatures with the support of machine learning techniques (Randhawa, 2020). This paper highlighted the major issues of covid-19 around 180 countries. They used combined method which includes machine learning methods and digital signal processing methods. This combined method is used for genome analysis, decision tree approach for augmented and result evaluation using Spearman’s rank correlation coefficient. This research work involved in many test analysis using classification methods. Each test contains different data values to identify the virus infection. The combined machine learning and digital signal processing methods are used to analysis the execution of different tests. The 10 fold cross validation is used for analyzing the outcomes.

WimNaude presented the effectiveness of artificial intelligence in the covid-19 prediction analysis (Wim Naudé., 2020). This paper also addressed the limitations, constrains and pitfalls. The artificial intelligence methods are used to provide the analysis of convid-19 spreading status in all over the world. The AI methods are required pre-training process for effective prediction. The Artificial intelligence methods are also used for fast prediction of covid-19 which can save human life. The used AI methods are effective to handle different kinds of data like, text data, image data and etc. The effective results of AI are discussed in this paper. The usages of Artificial Intelligence methods are helpful to handle the various types of data types.

AllaeErraissi et al presented a research work for the prediction of Covid-19 cases using machine learning methods (Allae Erraissi., 2020). The execution was done with the support of Spark ML software. With the help of this platform, data preparation, data enrichment and the development of algorithms were done.  This tool provides classification, regression, clustering and recommendation process. For this prediction process various machine learning methods were used such as decision tree, Gaussian Naïve Bayes, Support vector machine and Random forest tree. The partial differential equation methods were used for handling different parameters for this prediction process.

Yaohao Peng et al presented an overview of non-linearity and over fitting issues of machine learning in Covid-19 (YaohaoPeng., 2020). The support vector machine method was used for this research work. Five different structures of nonlinearity was tested with the use of five various kernel functions. The dataset was collected from 12 different countries. The sensitivity and prediction performance was measured with different hyper parameter settings. The linear kernel function was not effective in this prediction process and also it produced out-of samples forecasting. Support vector machine method was effective and produces improved results than the existing methods. The random forest and deep learning methods were not used in this research work. They also focused the LASSO and ridge regression methods were used with benchmark dataset. The kernel functions are highlighted and suggested for the future work for covid-19 prediction.

FurqanRustam et al proposed a future forecasting using supervised machine learning methods for the prediction of Covid-19 (FURQAN RUSTAM., 2020). The supervised learning methods are used for supporting the process of decision making which improves the prediction accuracy.  In this research work, the authors were focusing various machine learning techniques such as linear regression models, Least absolute shrinkage and selection operator, support vector machine and exponential smoothing. All these methods were used for predicting the infection of covid-19, death rates and recoveries rate. For this research work, they used three different types of prediction methods. These methods were executed for the prediction process for new case identification. Based on the execution results, the evaluation highlighted that least absolute and LASSO methods were more effective than the support vector machine.

Prabira Kumar Sethy et al proposed the deep features and support vector machine process for the prediction of corona virus disease (Prabira Kumar Sethy., 2020). The X-ray images are used for prediction process using SVM and deep features. The deep features are the extracted feature of fully connected layer process form Convolutional neural network and that was associated with support vector machine process. There are two prediction categories used in this research work. Those are pneumonia and normal. The prediction model consist SVM method with 13 number CNN models. The produced SVM method with deep features is effective in the prediction of Covid-19.  This model achieved 93.4% of accuracy. This was one of the highest accuracy in the prediction of Covid-19 using machine learning methods.

S.NoAuthorsPaper NameMethods usedOutcomesAccuracy
1.Narinder Singh Punn et alCOVID-19 Epidemic Analysis using Machine Learning and Deep Learning AlgorithmsDeep LearningUse Deep learning  models for pattern matching and predictionApproximately more than  90%
2.MucahidBarstugan et alCoronavirus (COVID-19) Classification using CT Images by Machine Learning MethodsSupport Vector MachineFind the maximum classification accuracy using CT images.99.68%
3.Mohammad Pourhomayoun et alPredicting Mortality Risk in Patients with COVID-19 Using Artificial Intelligence to Help Medical Decision-MakingNeural network, Random forest, SVM, Decision Tree, K-NN and Logistic Regression methods are usedNeural network was effective when compared with other methods in prediction of Covid-1993%
4.Muzafar Bhat et alSentiment analysis of social media response on the Covid19 outbreak

 

Sentiment analysisTwitter messages were collected for sentiment analysisMost of People reacted as positive sentiment
5.SwapnarekhaHanumanthuRole of Intelligent Computing in COVID-19 Prognosis: A State-of-the-Art Review

 

Deep learning methods and Machine learning methodsAnalyzed the performance of Random forest, SV, linear regression, CNN model, RNN modelsSVM produced 97.48% accuracy
6.Jiangpeng Wu et. alRapid  and  accurate  identification  of  COVID-19  infection  through  machine

learning based on clinical available blood test results

 

Random Forest with discrimination toolCOVID-19 related blood samples were extracted and analyzed for prediction95.95%

 

Chapter-3 Additional Section

Project Planning

The project is done for the Covid ‘19 Sentimental analysis. It includes several phases like literature review, system design and development, implementation, Testing and deployment, Maintenance and finally comes with document preparation. In the Literature review several papers are surveyed and knowledge is gained how the covid analysis is done in different techniques and performance with accuracy is also analyzed. Next it comes with the system design and data collection where the dataset is collected and its attributes are analyzed. Design and development process is carried out to analyze the requirements in the system. The specification samples are collected and the analysis is carried out. The next phase is implementations in which the mat lab software is used for implementation process. Various plots are generated and results are analyzed. Once the implementation process is done maintenance is done with necessary modifications based on the functionalities. Finally the document is prepared according to the given template.

The literature review is done based on the sentimental analysis in which the Covid prediction is done. Several researches are going on to predict the family of virus and its spread across the globe. Sentimental analysis is used for the covid prediction in which literature is done for 3 weeks by analyzing the different concepts. System design and data collection is next process in which the data can be collected from kaggle in which the analysis is done. Analysis process is done with different techniques using sentimental analysis and the prediction is done. System design is complete in 3 weeks with the analysis process. Implementation of the project is done with mat lab in 3 weeks and the various results are predicted with sentimental analysis. Graphs are generated to show the efficiency and accuracy. Next process is testing process in which the various testing process is done. Finally the document is prepared with the template given by the university with 6 chapters and some appendices.The researchers are doing further research to stop the wide spread of corona virus.

The Gannt chart is created based on the dates allocated for the project starting from collecting requirements till the document preparation. The document is prepared with 6 sections in which it includes all the necessary requirements and specification.

S.NoTask NameDuration In days
1Literature Review20
2System Design and Data collection15
3Implementation of the project20
4Testing and evaluation process10
5Maintenance of the system10
6Document preparation15

 

The Gann’t chart is done using excel showing each phase of the covid analysis.Machine learning methods were used for accurate prediction of Covid-19 using the blood test result. 49 clinical available blood samples are used for prediction of Covid-19. These blood samples were extracted using random forest method. By using this method 95.95% of accuracy is reached. The Discriminative tool was used for execution of Covid-19 prediction. The proposed method also achieved 95.12% of sensitivity and 96.96% of specificity by using the blood sample data. The accuracy and specificity is calculated based on the dataset selected.

The sample collected can be examined for the virus and the necessary medication is given. Further process can be done in the maintenance phase if needed. The data collection is done by the user which takes more time. Here the data is extracted and the analysis is done using matlab.Results are generated and shown below. The data can be analyzed with different techniques to improve the accuracy and efficiency. The data analysis process can be done with different samples collected in the real time. The performance results were compared with the accuracy, sensitivity and specificity produced by random forest, Support vector machine, K-NN, Logistic Regression, Convolution Neural Network (CNN) and Recurrent Neural Network (RNN). Among all these methods, the Support Vector machine model produced high accuracy in prediction of Covid-19.This model achieved 97.48% classification accuracy.

Chapter-4

Code

 

The chapter-3 contains the MATLAB code for prediction of COVID-19 using neural network and Support Vector Neural Network methods with sentimental analysis. With the help of extracted data from the UCI repository, analysis of COVID-19 is performed. The emotion of COVID-19 is predicted by sentimental analysis process and the classification of data is done by using support vector machine. The covid-19 prediction was done by using the MATLAB environment. This MATLAB supports the implementation of machine learning methods.

The support vector neural network method is used the pre-training methods that are helpful to train the model for making appropriate predictions. The dataset is divides into two parts for training and testing. The maximum dataset is used for training to avoid over fitting complexities.  The training method includes the standard squared error, standard square weight and the effectiveness of used parameters in prediction.

The neural network methods are used to calculate the sum square error and the Bias values for prediction. The effective calculation includes error surface including weights and error contour using Bias of weights.

Then the sentiment analysis process is done to find the emotions about COVID-19. The classification method is used to classify the messages based on the classes. The steps involved for the proposed framework is  collection of data from the repository of UCI ,the data is preprocessed ,sentiment analysis is performed to find the positive and negative emotion, the case is classified ,data frames are created with results,dataframes are sorted, the comparison for the result accuracy is predicted using the SVNN and simple NN and results are produced using ,methods of visualization.

Code

 

Code using Support vector Neural Network

function [W1,W2,b1,b2] = SVNN(X,y,nneu)

C=70;

max_indiv = 1500;

pressao = 5;

elitism = 1;

max_resol = 12;

iter_max = 10;

 

[L,namost] = size(X);

Positives = [];

Negatives = [];

for k=1:namost

if y(k)==1

positives = [positives,k];

else

negatives = [negatives,k];

end

end

sp = max(size(positives));

sn = max(size(negatives));

maxim = max(max(X));

minem = min(min(X));

 

The MATLAB code is implemented for the prediction of Covid-19 using a new training proposed method Support Vector Neural Network and Simple Neural Network. SVNN is a non-probabilistic model that can be used for representation of text. The new texts are classifies using the similarities. From the code The MATLAB SVNN function is used for the implementation of SVNN.The input argument are given in a matrix. From the matrix X is the column vector and Y is the row vector. The hidden neutron parameter nneu is used .The elements are the based on own target classes should be -1 or 1.The MLP parameter for the algorithm W1,W2,b1,b2 these are argument of input for the simulator MLP.The MLP simulator involves the testing data matrix and the target vector. After the declaration of function of SVNN, the setting of parameter is performed. Set C is 70 as the punishing parameter and set the initial population is 1500.The selective pressure is placed as 1 to 8 .The setting of Elitism 1 is yes and 0 is no. The setting of generation per resolution is specified. By using the class separation is done and then population is generated. Then the generation of loop is performed and ranking of individual calculated. The evaluation is done with dataset for generating the error. The next step is the training dataset of negative is reduced to deal with the dataset which is unbalanced. The testing of positive and negative result is performed for the evaluation of Covid-19.The population ranking of fitness is performed for the ranking the results.

Code Using Simple Neural Network

The connected fully simple neural network is used for the prediction of Covid-19.The simple neural network methods are used to make some random predictions that are matched with correct output. In this process the error values are considered to check the efficiency of the algorithm. The difference between the produced values and the actual values are calculated. Training process is required for this simple neural network to make predictions. The random number of hidden layers, functions for activation, parameters and sets for validation was taken for the use. The simple neural network is designed for classification and y value are labeled as integer and number for each is match to one class.

function [acc,estimated] = sim_NN(W1,W2,b1,b2,X,t)

[L,Col] = size(X);

Error = 0;

Estimated = zeros(1,Col);

for k = 1:Col

estimated(k) = W2*logsig(W1*X(:,k)+b1)+b2;

estimated_aprox = sign(estimated(k));

if not(estimated_aprox==t(k))

error = error+1;

end

end

acc = (Col-error)/Col;

 

From the code The MATLAB sim_NN function is used for the implementation of simple neural network .The input argument are given in a matrix. From the matrix X is the column vector and the MLP parameter for the algorithm W1, W2, b1,b2,t these are argument of input for the simulator MLP.The MLP simulator involves the testing data matrix and the target vector. The error is set to 0.The estimation of weight for neuron is calculated using the weight calculating functions.

The simple neural network method is used to calculate the sum square error and the Bias values for prediction. The MATLAB code for simple neural network for the prediction of Covid-19 for the sentimental analysis and prediction of result. The parameter for the neuron and input vector is defined. The output of neuron is calculated. The neuron output is plotted over the input range. From the connected neural network basic training and prediction function is first defined. The random number of hidden layer is used. The results produced the effective calculation includes error surface including weights and error contour using Bias of weights. Based on the training, the algorithms are able to predict the disease. The simple neural network methods are used to calculate the sum square error and the Bias values for prediction. The effective calculation includes error surface including weights and error contour using Bias of weights.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Chapter-5

Evaluation

The prediction methods are based on the neural network and support vector neural network methods.

The execution is done with the help of extracted data from the UCI repository. The text processing methods will be implemented to remove the stop words. Then the sentiment analysis process will be done to find the emotions about COVID-19. The classification methods will be used to classify the messages based on the categories.

The framework includes the following steps:

  • Step-1: Data collection form UCI Repository
  • Step-2: Text Pre-processing
  • Step-3: Perform Sentiment Analysis
  • Step-4: Classify the case
  • Step-5: Create data frame with results
  • Step-6: Sort Data frame
  • Step-7: Accuracy comparison
  • Step-8: Result Producing

Visualization methods will be used to produce the results in an effective manner.

4.1 Dataset Details:

The dataset is derived from Kaggle repository. This dataset contains the data about the convid-19 patients in South Korea. This dataset contains 8 different attributes such as patient ID, global number, and sex, year of birth, province, city and disease. This disease includes various factors such as infection case, infection order, the ID of infected patient, contact number of the patient, symptom onset date, conformed date, release date, deceased date and the state of the patient. This state includes three different specifications like whether the patient is taking treatment with isolation state, released or deceased. All these attributes are stored in the taken dataset. This also contains 3254 records. The whole dataset is used for this prediction process.  URL: https://www.kaggle.com/sudalairajkumar/novel-corona-virus-2019-dataset#.This dataset contains two different file. Those are death dataset and recovered dataset. The death dataset contains the data of death due to this Convid-19. The recovered dataset specifies the data of recovered persons. The COVID-19 data set is inadvertently proving to be a super-interesting pragmatic test for AI based analysis said Anthony Goldbloom chief executive of the Kaggle website.

 

 

Fig 4.1 Dataset details: Death Dataset

Fig No: 4.2 Dataset Details: Recovered Dataset

Methodology:

4.2 Simple Neural network

Neural network method are used for solving various problems in different sectors like business sectors, healthcare sectors and etc. this methods is used to produce effective results in sale forecasting, validation of data and risk management.  This is one of the supervised learning algorithms that provide input data which contains independent variables. The simple neural network methods are used to make some random predictions that are matched with correct output. In this process the error values are considered to check the efficiency of the algorithm. The difference between the produced values and the actual values are calculated. Training process is required for this simple neural network to make predictions. Based on the training these methods, the algorithms are able to predict the disease.

Neural network is a network of neurons generated by artificial nodes to solve different artificial intelligence problems. This is a predictive modeling approach which is used for adaptive control and applications. These are trained with the support of datasets. This simple neural network model contains two input nodes and one output node to produce the result.  This neural network has strong processor that supports the arrangement the neural network process as tiers. The first tier collects the raw information and sends it to the hidden nodes through input nodes.

Fig no: 4.3 Simple Neural Networks

 

4.2.1 Generation of Simple Neural Network

The following steps are used to construct a simple neural network.

Step 1: The input nodes are connected to produce the output through output nodes.

Step 2: Three weight parameters are used to connect the input nodes for processing

Step3:  The information can be transferred in two ways.

Step 4: Pass the results to activation function

Step 5: Produce the results via output nodes.

4.3 Types of Neural Networks

In simple, this neural network method is classified into Different types. Those are

  • Feed Forward Neural Network
  • Radial Basis Function Neural Network
  • Multilayer Perceptron
  • Convolutional Neural Network
  • Recurrent Neural Network
  • Modular Neural Network

These methods are used based on the requirement. All these neural network methods has three basic type of layer processing. Those are input layer, hidden layer and output layer processing. These different types are used to achieve the results in different ways. Based on the selection of neural network types, the outcomes will differ.

4.3.1. Feed Forward Neural Network

This feed forward neural network method is used for predicting the case by using sequential learning process. This method is not a time dependent method. This method supports gradient based learning method for training the algorithm. This method mainly supports the stochastic gradient decent method for minimizing the cost function. Adagrad, Adam and RMSProp methods are the other learning methods used for optimization.

4.3.2. Radial Basis Function Neural Network

This is a type of neural network which used the radial function for processing. This function is used as the activation function in the neural network method. This type is used to produce the values in association with function approximation, prediction of time series and classification.

4.3.3. Multilayer Perceptron

This is a supervised learning method that supports the process of back propagation training to produce the improved results. The multiple layer concepts are used in this process and non-linear activation is processed. This method can use either hyperbolic tangent method or logistic function for activation process.

4.3.4. Convolutional neural network

This convolutional neural network is used to reduce the dimensionality and this method supports the feature map concept to select the effective method for prediction. There are four different layers are in this process. Those are convolutional layer, pooling layer, fully connected layer and activation function. All these layers are used to produce the effective results based on the pre-training process given to the algorithm.

4.3.5. Recurrent Neural Network

This model contains memory that supports to remember the information which are used for calculation. The same parameters are used for input values and that helps to perform the operation.  This method helps to reduce the difficulties of using the parameters in calculation. This method also supports the independent activation function that helps to reduce the complexity of using increased number of parameters. The main advantage of using this method is the support of remembers all the information throughout the process.

4.3.6. Modular Neural Network

This model consist series of neural network that are independent from one another. These independent neural networks can use separate input values and process for producing the output. These processes are considering as subtask of the neural network and finally the best output will be produced. To collect all the output from subtask, the intermediary is used. The main consideration of this model is the modules in this network do not have ant interaction process with each other.

4.4 Support Vector Neural Network

This support vector neural network method was proposed by O.Ludwig. This author used this method for the face recognition process using neural network concept. This model supports both support vector machine methods and neural network method for prediction. This integrated method is useful to produce the improved result in the selected problems. The main advantage of using this support vector machine method is producing the results for both classification and regression problems. This method is more effective to produce the classification results using class labels and also produce the numerical values for the result of regression problems. This method supports the clear margin for making separation of data for classification. This algorithm can be implemented using kernel which is used to transform the input values into the required form. This is one of the supervised learning methods which required training process for making classification. This support vector machine method is highly effective while using the high dimensional data. This also handles the memory efficient methods that can support the high dimensionality processes. Different kernel functions are supported by this support vector machine that can specify the decision function.

4.5 Sentiment analysis using machine learning

The sentiment analysis is one of the machine learning methods which is used to produce the two different opinions like positive opinion and negative opinion. This sentiment analysis is used to understanding the people emotion. This sentiment analysis is used for various sectors like understanding the customer’s choice in market products, their opinion about the product and their requirements. All these analysis are useful for business to provide useful products that are required by the customers. This sentiment analysis is providing the feedback of customers and people in social media.

4.5.1 Types of sentiment analysis

There are different types of sentiment analysis process is done to understand the feelings of people. Those are

  • Fine-grained sentiment analysis
  • Emotion detection
  • Aspect-based sentiment analysis
  • Multilingual sentiment analysis

Fine-grained sentiment analysis:

This method is used to produce the results in described categories such as very positive, positive, negative and very negative. These categorizations are analyzed by the stars given by the customers. The very positive category indicates the maximum star values and the very negative category indicates the minimum star values denoted by the customers.

Emotion detection

This detection is used for finding the different feelings of people like happy, sad, anger, fear, frustration and etc. this emotion detection is done by using lexicons or machine learning algorithms with complex methods. The lexicons based detection is done by the words used by the people.

Aspect based sentiment analysis:

This aspect based sentiment analysis is used to categorize the data into three different categories such as positive, neutral and negative. This analysis also provides the review of these classes based on the selection of words.

Multilingual Sentiment analysis:

This is one of the complex analyses which include more number of processes. This analysis consist the data resources that are available in online.

4.5.2. Working methodology of Sentiment analysis

The working process of sentiment analysis includes two different phases. Those are

  • Training Phase
  • Prediction phase

The training process includes the process of selecting the tags and feature extraction methods to select the particular input for getting the appropriate output. The feature extractor is done the process of text conversion which helped to transfer the text into vector for output prediction. The feature tags are inputted into the machine learning algorithms that are helpful for building a prediction model.

Fig No: 4.4 Sentiment analysis – workflow

The feature extraction methods are used to transfer the words into ngrams for finding the class of the messages. After selecting the features the classification algorithms are implemented to find the class using the predefined tags.

4.5.3. Classification algorithms:

The classification algorithms are used to classify the record into the appropriate classes using class labels. For this sentiment analysis some famous classification methods are used. Those are naïve Bayes, Linear Regression, Support vector machine and deep learning.

Naïve Bayes: This classification method is used to classify the text based on the Bayes theorem. The effectiveness of this method is discussed by various researchers.

Linear Regression: This linear regression method is used to predict the values using the given features.

Support Vector Machine: This is a non-probabilistic model that can be used for representation of text. The new texts are classifies using the similarities.

Deep learning: this method supports different classification algorithms that are based on the human brain working process. This model is based on the artificial neural network model.

4.6 Execution results:

The covid-19 prediction was done by using the MATLAB environment. This MATLAB supports the implementation of machine learning methods.

 

Fig No: 4.5 weight vector selection

This picture indicates the selection of values based on the weights. The simple neural network method is used to provide the method of calculating and selecting the weighted valued for the network model.

Fig No: 4.6 Training process using dataset

The support vector neural network method is used the pre-training methods that are helpful to train the model for making appropriate predictions. The dataset is divides into two parts for training and testing. The maximum dataset is used for training to avoid over fitting complexities.  The training method includes the standard squared error, standard square weight and the effectiveness of used parameters in prediction.  The given results produced that the SSE values is 0.116236. The Standard square weight is 1649.18. The effectiveness of training is calculated using 100 epochs. For 100 epochs, the produced result is 28.237. From this training the selected support vector neural network method will be effectively predicting the cases in Covid-19 dataset.

Fig No: 4.7 Error Surface and Error Contour

The neural network methods are used to calculate the sum square error and the Bias values for prediction. The given execution results produced the effective calculation includes error surface including weights and error contour using Bias of weights.

Fig no: 4.8 Results analysis- Accuracy Prediction

From this result, the accuracy for two different files are produced. The accuracy for the death file is 97.64%. For the recovered file, the accuracy is reached as 87%.

To calculate the accuracy, sensitivity and specificity following formula is used.

By using this formula, the accuracy of the support vector neural network method is calculated. This also provides good results in the prediction of covid-19 by providing effective values in sensitivity and specificity values.

Numerical Results

TPTNFPFNAccuracyTPRTNR
Deaths74935870035516548154981695795.6797.6451.64
Recovered12826815100049161154413102614564134187.0093.9852.94

 

This table Indicates the True Positive, True Negative, False Positive and False Negative values to calculate the Accuracy, sensitivity, specificity for two different file used in this Covid-19 prediction process. The Testing accuracy of death file using support vector neural network is 95.67% and the recovered file is 87%. This is one of good prediction accuracy which is achieved using the support vector neural network method.

Chapter-6

Conclusion & Future scope

This examination work centers the sentimental analysis of COVID-19 on the globe. Social information on web contains some genuine occasions, for example, COVID-19 and so on. The sentiment analysis is one of the machine learning methods which is used to produce the two different opinions like positive opinion and negative opinion. This sentiment analysis is used to understanding the people emotion. This sentiment analysis is used for various sectors like understanding the customer’s choice in market products, their opinion about the product and their requirements numerous individuals including government associations can hear the point of view of different individuals through this online media. In light of feeling investigation the realities about the infections, the spreading proportion and the familiarity with this malady can be shared by various individuals all around the globe. This feeling investigation can be valuable for getting the client mindfulness and their activities against COVID-19. Cloud computing is one of the ground-breaking innovation which can be utilized to store the web-based media information like twitter information, face book shares and so on. Tweets and messages were analyzed which are shared by different clients can be utilized to discover various feelings identified with this sickness. The dataset is derived from Kaggle repository. This dataset contains the data about the convid-19 patients in south Korea. This dataset contains 8 different attributes such as patient ID, global number, sex, year of birth, province, city and disease. This disease includes various factors such as infection case, infection order, the ID of infected patient, contact number of the patient, symptom onset date, conformed date, release date, deceased date and the state of the patient. This dataset contains two different file. Those are death dataset and recovered dataset. The death dataset contains the data of death due to this Convid-19. The recovered dataset specifies the data of recovered persons. We tended to issues encompassing open feeling reflecting profound worries about Corona virus, prompting the distinguishing proof of development in dread assessment and negative estimation. Likewise we showed the utilization of exploratory and elucidating literary investigation and analysis of textual data, to find beginning phase experiences, for example, by gathering of words by levels of a particular non-text variable. At last, we gave a correlation of literary grouping systems utilized in man-made reasoning applications, and exhibited their value for shifting lengths of Tweets. Along these lines, the current investigation gave techniques important enlightening and open opinion experiences age potential, which can be utilized to grow truly necessary inspirational arrangements and methodologies to counter the fast spread of news related to Corona virus. The covid-19 prediction was done by using the MATLAB environment. This MATLAB supports the implementation of machine learning methods.

 

Given the simple accessibility of COVID-19 related huge information, a broad cluster of investigation and best in class AI driven arrangements should be created to address the pandemic’s worldwide data complexities. While the ebb and flow research stream adds to the vital cycle, much more should be done over numerous web-based media, news and open and individual correspondence stages. Such arrangements will likewise be basic in recognizing a supportable pathway to recuperation post corona for instance, understanding open viewpoints and slant utilizing printed examination and AI will empower strategy producers to take into account open needs more explicitly and furthermore plan estimation explicit correspondence techniques. Organizations and independent ventures can likewise profit through such examinations and AI models to more readily comprehend buyer feeling and desires. Future examinations can glance in to pre and post lockdown tweets and comprehend whether there was an adjustment in estimations from the earliest starting point to the furthest limit of the lockdown. Likewise, future investigations can glance in to factors that influence metal wellbeing during lockdowns and pandemic spreads. Another zone for future exploration could be handling of phony news that gets coursed through online media, affecting the emotional well-being of the beneficiaries.

 

 

References

Thanh Thi Nguyen., 2020. Artificial Intelligence in the Battle against Corona virus(COVID – 19): A Survey and Future Research Directions. Research gate, p. 14.

Samir Kumar Bandyopadhyay Sr., S. D. J., 2020. Machine Learning Approach for Confirmation of COVID-19 Cases: Positive, Negative, Death and Release. medRxiv.

ShreshthTuli., S. R. S., 2020. Predicting the growth and trend of COVID-19 pandemic using machine learning and cloud computing. Internet of Things, pp. Vol-11.

MilindYadav., M. D., 2020. Analysis on novel coronavirus (COVID-19) using machine learning methods. Chaos, Solitons & Fractals, pp. Vol-139.

Biraja Ghoshal, A. T., 2020. Estimating Uncertainty and Interpretability in Deep Learning for Coronavirus (COVID-19) Detection. arXiv.org.

Xiang Bai., C. F. Y. Z. S. B. Z. L. L. X. Q. C. Y. X. T. X. S. G. X. X. D. S. R. D. C. Z. C. C. D. N. L. Q. W. C., 2020. Predicting COVID-19 Malignant Progression with AI Techniques. SSRN.

Randhawa, G. S. S. M. P. M. E. R. H. d. S. C. P. E. H. K. A. &. K. L., 2020. Machine learning using intrinsic genomic signatures for rapid classification of novel pathogens: COVID-19 case study. PLOS ONE, pp. Vol-15, issue-4.

Wim Naudé., 2020. Artifcial intelligence vs COVID‑19: limitations, constraints and pitfalls. AI & Society.

Allae Erraissi., M. A. A. B. M. B., 2020. Machine Learning model to predict the number of cases contaminated by COVID-19. Research Square.

YaohaoPeng., M. H., 2020. An empirical overview of nonlinearity and overfitting in machine learning using COVID-19 data. Chos Solitons & Fractals, pp. Vol-139.

FURQAN RUSTAM., A. A. R. A. M. S. U. O. W. A. G. S. C., 2020. COVID-19 Future Forecasting Using Supervised Machine Learning Models. IEEE Access.

Vinay Chamola., V. H. G. G., 2020. A Comprehensive Review of the COVID- 19 Pandemic and the Role of IOT, Drones, AI, Blockchain and 5G in managing its Impact. IEEE, p. 41.

Prabira Kumar Sethy., S. K. B. P. K. R. P. B. .., 2020. Detection of coronavirus Disease (COVID-19) based on Deep Features and Support Vector Machine. International Journal of Mathematical, Engineering and Management Sciences, pp. Vol-5, no-4, pp 643-651.

Nasir Saeed., A. B. Y. A. N. S. A., 2020. Wnen Wireless Communication Faces COVID-19: Combating the Pandemic and Saving the Economy. Arxiv, p. 11.

Rucha Visal ., A. T. S., 2020. Monitoring Social Distancing for Covid-19 Using OpenCV and deep learning. International Research Journal of Engineering and Technology, p. 3.

Lalmuanawma, S. H. J. &. C. L., 2020. Applications of Machine Learning and Artificial Intelligence for Covid-19 (SARS-CoV-2) pandemic: A review.. Chaos, Solitons & Fractals,.

Alimadadi, A. A. S. M. I. M. P. B. J. B. &. C. X., 2020. Artificial Intelligence and Machine Learning to Fight COVID-19. Physiological Genomics.

Parul Arora., H. K. B. K. P., 2020. Prediction and analysis of COVID-19 positive cases using deep learning models: A descriptive case study of India. Chaos, Solitons and Fractals.

Gergo Pinter., I. F. A. M. P. G. R. G., 2020. COVID-19 Pandemic Prediction for Hungary;A Hybrid Machine Learning Approach. Mathematcis, pp. Vol-8, issue-6..

Amir Ahmad., S. G. S. K. R. G. K. S. J. M. &. O. M. B., 2020. The Number of Confirmed Cases of Covid-19 by using Machine Learning: Methods and Challenges. Archives of Computational Methods in Engineering.

Allae Erraissi., M. A. A. B. M. B., 2020. Machine Learning model to predict the number of cases contaminated by COVID-19. Research Square.

Almatarneh, S. & Gamallo, P., 2019. Comparing supervised machine learning strategies and linguistic features to search for very negative opinions.

Amir Ahmad., S. G. S. K. R. G. K. S. J. M. &. O. M. B., 2020. The Number of Confirmed Cases of Covid-19 by using Machine Learning: Methods and Challenges. Archives of Computational Methods in Engineering.

Bhat R., S. V. N. N. K. C. M. P. K. N., 2020. COVID 2019 outbreak: the disappointment in Indian teachers. Asian J. Psychiatry..

Biraja Ghoshal, A. T., 2020. Estimating Uncertainty and Interpretability in Deep Learning for Coronavirus (COVID-19) Detection. arXiv.org.

Boldog P., T. T. V. Z. D. A. B. F. R. G., 2020;. Risk assessment of novel coronavirus COVID-19 outbreaks outside China. J. Clin. Med.. p. 9(2):571.

Chen, X. et al., 2020. Trends and Features of the Applications of Natural Language Processing Techniques for Clinical Trials Text Analysis.

El Zowalaty M.E., J. J., 2020. From SARS to COVID-19: a previously unknown SARS-CoV-2 virus of pandemic potential infecting humans – call for a one health approach. One Health.

FURQAN RUSTAM., A. A. R. A. M. S. U. O. W. A. G. S. C., 2020. COVID-19 Future Forecasting Using Supervised Machine Learning Models. IEEE Access.

Gergo Pinter., I. F. A. M. P. G. R. G., 2020. COVID-19 Pandemic Prediction for Hungary;A Hybrid Machine Learning Approach. Mathematcis, pp. Vol-8, issue-6..

Goyal K., C. P. C. K. G. P. S. M. .., 2020. Fear of COVID 2019: first suicidal case in India. Asian J. psychiatry. .

Jiangpeng Wu., P. Z. L. Z. W. M. J. L. ,. C. T. ,. Y. L. ,. J. C. ,. Z. Y. ,. J. Z. M. Z. H. H. X. X. S. L., 2020. Rapid and accurate identification of COVID-19 infection through machine learning based on clinical available blood test results. medrxiv.

Jin, D., Jin, Z., Zhou, J. & Szolovits, P., 2019. . Is bert really robust? natural language attack on text classification and entailment.

Kowsari, K. et al., 2019. Text classification algorithms: A survey..

Kretinin, A., Samuel, J. & Kashyap, R., 2018. When the Going Gets Tough, The Tweets Get Going! An Exploratory Analysis of Tweets Sentiments in the Stock Market.

Lalmuanawma, S. H. J. &. C. L., 2020. Applications of Machine Learning and Artificial Intelligence for Covid-19 (SARS-CoV-2) pandemic: A review.. Chaos, Solitons & Fractals,.

MilindYadav., M. D., 2020. Analysis on novel coronavirus (COVID-19) using machine learning methods. Chaos, Solitons & Fractals, pp. Vol-139.

Mohammad Pourhomayoun., M. S., 2020. Predicting Mortality Risk in Patients with COVID-19 Using Artificial Intelligence to Help Medical Decision-Making. medRxiv.

Mucahid Barstugan., U. O. S. O., 2020. Coronavirus (COVID-19) Classification using CT Images by Machine Learning Methods. Research Gate.

Muzafar Bhat., M. Q. N.-u.-A. B. M. K. N. A. B. A., 2020. Sentiment analysis of social media response on the Covid19 outbreak. Elsevier Public Health Emergency Collection.

Narinder Singh Punn, S. K. S. S. A., 2020. COVID-19 Epidemic Analysis using Machine Learning and Deep Learning Algorithms. medRxiv.

Nasir Saeed., A. B. Y. A. N. S. A., 2020. Wnen Wireless Communication Faces COVID-19: Combating the Pandemic and Saving the Economy. Arxiv, p. 11.

Naudé., W., 2020. Artifcial intelligence vs COVID‑19: limitations, constraints and pitfalls. AI & Society.

Nguyen., T. T., 2020. Artificial Intelligence in the Battle against Corona virus(COVID – 19): A Survey and Future Research Directions. Research gate, p. 14.

Parul Arora., H. K. B. K. P., 2020. Prediction and analysis of COVID-19 positive cases using deep learning models: A descriptive case study of India. Chaos, Solitons and Fractals.

Prabira Kumar Sethy., S. K. B. P. K. R. P. B. .., 2020. Detection of coronavirus Disease (COVID-19) based on Deep Features and Support Vector Machine. International Journal of Mathematical, Engineering and Management Sciences, pp. Vol-5, no-4, pp 643-651.

Rajuj Vaishya, M. J. I. H. K. A. H., 2020. Artificial Intelligence (AI) applications for COVID-19 pandemic. Diabetes & Metabolic Syndrome: Clinical Research & Reviews, pp. 337-339.

Randhawa, G.. A. &. K. L., 2020. Machine learning using intrinsic genomic signatures for rapid classification of novel pathogens: COVID-19 case study. PLOS ONE, pp. Vol-15, issue-4.

Rocha, G. & Lopes Cardoso, H. 2., 2018. Recognizing textual entailment: Challenges in the Portuguese language. Information.

Rucha Visal ., A. T. S., 2020. Monitoring Social Distancing for Covid-19 Using OpenCV and deep learning. International Research Journal of Engineering and Technology, p. 3.

  1. Oh, S. L. a. C. H., 2020. . “The Effects of Social Media Use on Preventive Behaviors during Infectious Disease Outbreaks: The Mediating Role of Self-relevant Emotions and Public. Risk Perception,” Health Communication, , pp. pp. 1-10,.

Samir Kumar Bandyopadhyay Sr., S. D. J., 2020. Machine Learning Approach for Confirmation of COVID-19 Cases: Positive, Negative, Death and Release. medRxiv.

ShreshthTuli., S. R. S., 2020. Predicting the growth and trend of COVID-19 pandemic using machine learning and cloud computing. Internet of Things, pp. Vol-11.

Skoric, M., Liu, J. & Jaidka, K., 2020, 11, 187. Electoral and Public Opinion Forecasts with Social Media Data: A Meta-Analysis. Information.

Swapna rekha Hanumanthu., 2020. Role of Intelligent Computing in COVID-19 Prognosis: A State-of-the-Art Review. Elsevier Public Health Emergency Collection.

Vijayan, V., Bindu, K. & Parameswaran, L. A., 2017. comprehensive study of text classification algorithms. In Proceedings of the 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI), Udupi, India, ;.

Vinay Chamola., V. H. G. G., 2020. A Comprehensive Review of the COVID- 19 Pandemic and the Role of IOT, Drones, AI, Blockchain and 5G in managing its Impact. IEEE, p. 41.

Wang H., W. Z. D. Y. C. R. X. C. Y. X. W. Y., 2020. Phase-adjusted estimation of the number of coronavirus disease 2019 cases in Wuhan, China.

Wim Naudé., 2020. Artifcial intelligence vs COVID‑19: limitations, constraints and pitfalls. AI & Society.

Xiang Bai., C. F. Y. Z. S. B. Z. L. L. X. Q. C. Y. X. T. X. S. G. X. X. D. S. R. D. C. Z. C. C. D. N. L. Q. W. C., 2020. Predicting COVID-19 Malignant Progression with AI Techniques. SSRN.

YaohaoPeng., M. H., 2020. An empirical overview of nonlinearity and overfitting in machine learning using COVID-19 data. Chos Solitons & Fractals, pp. Vol-139.

 

 

 

 

Leave a Comment