ABSTRACT
The pandemic Covid-19 is a name coined by WHO on 31st December 2019. This devastating illness was carried on by a new coronavirus known as SARS-COV-2. Most of the research has focused on estimating the total number of cases and mortality rate of COVID-19. Due to this, people across the world were stressed out by observing the growing number of cases every day. As a means of maintaining equilibrium, this paper aims to identify the best way to predict the number of recovered cases of Coronavirus in India. Dataset was divided into two parts: training and testing. The training dataset utilised 70% of the dataset, and the testing dataset utilised 30%. In this paper, we applied 10 machine learning techniques i.e. Random Forest Classifier (RF), Naive Bayes (NB), Quadratic Discriminant Analysis (QDA), Gradient Boosting Classifier (GBM), Linear Discriminant Analysis (LDA), Logistic Regression (LR), K Neighbour Classifier (KNN), Decision Tree Classifier (DT), SVM - Linear and Ada-Boost Classifier in order to predict recovered patients in India. Our study suggests that Random Forest Classifier outperforms other machine learning models for predicting the recovered Coronavirus patients having an accuracy of 0.9632, AUC of 0.9836, Recall of 0.9640, Precision of 0.9680, F1 Score of 0.9617 and Kappa of 0.9558. © 2022 IEEE.
ABSTRACT
Machine learning for Covid-19 diagnosis from blood tests is a topical problem. Many studies of this problem are mainly devoted to comparing various algorithms' efficiency. However, the first and often the most critical part of machine learning is the preparation of a relevant and correct dataset of the required size for developing the generalization models. This study demonstrates the lack of the models' generalization performance based on some publicly available datasets. That leads to the futility of such models in practice even if they were developed using the best algorithms and achieved high metrics. Therefore, another dataset is proposed. Its features are discussed. This dataset splits into training and testing sets by stratification due to an imbalanced data structure. Machine learning models of the problem by various algorithms are developed based on the proposed dataset. The modelling results on the testing set have demonstrated that the best models - Gradient Boosting Classifier with fixing imbalance methods SMOTE and ADASYN, TensorFlow and Gene Expression Programming - handle negative Covid-19 diagnosis well enough since they have high precision and high recall. However, mixed signals have been obtained for a positive Covid-19 diagnosis. TensorFlow and Gene Expression Programming models have high precision and relatively low recall for positive Covid-19 diagnosing. It means these models can't detect Covid-19 well enough but are highly reliable when they do. Gradient Boosting Classifier models do not have enough high precision and recall for positive Covid-19 diagnosing. New challenges of machine learning for Covid-19 diagnosis based on blood tests are found for future work. © 2022 IEEE.
ABSTRACT
Coronavirus (COVID-19) had major impacts on the daily lives of people. Lock-downs, work from home situations, loss in jobs, market changes, and less communication, and interaction between people especially during the stressful Covid period have made them more vulnerable to mental health issues, depression, loneliness, etc. With Covid related healthcare being given priority, the mental health issues faced by the public that has been both directly and indirectly affected by it have been majorly left ignored. These issues need to be taken care of by people on individual level and by the government for better public health. Hence, in this paper we introduce the emerging technique of data mining into the Covid-19 linked mental health for predicting the susceptibility of the general public around the globe to mental health side effects as a result of covid and pandemic circumstances. We used the COVIDiSTRESS survey data containing 103825 instances of people across the globe to identify the people more susceptible to Covid related stress. Logistic regression, random forest, xgboost, AdaBoost, and gradient boosting classifier were applied to the processed data giving an accuracy of 88.12%, 88.89%, 88.73%, 88.60%, and 89.25% respectively. The Models predicted the people who are likely to face covid stress based on different independent factors like their demographic variables, trust of authorities, corona concerns etc. The stress factor was measured using PSS-10 variable included in the survey. The result showed that the model developed with Gradient Boosting Classifier is found to be the most efficient model with an accuracy of 89.25%. Our analysis also showed that females, divorced/widowed people and full-time employees were more prone to stress amongst others in the gender/marital status/employment category. © 2022 IEEE.
ABSTRACT
The importance of social media has seen a huge growth in last few years as it lets folks from any part of the world remain in contact. Due to COVID-19 epidemic, social media is used extensively and has turned to be more pertinent than the past years. Alongside, fake news dissemination has revived and also dissemination of tweets has taken the attention. For the present study, we used various machine learning models to detect the fake news on COVID-19 related tweets. Due to millions of active social media users, identifying the fake news has become a crucial task. The models we applied are Gradient Boosting Classifier (GBC), Logistic Regression (LR), Random Forest Classifier (RFC), Decision Tree Classification (DT). All these models detect if the tweet relating to COVID-19 is “Fake News” or “Not a Fake News”. Hence we conclude that Logistic Regression is best among all other models. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.