Identification of Insecurity in COVID-19 Using Machine Learning Techniques

Abstract

During the pandemic COVID-19, many people died due to the infection caused by the deadly virus, and many affected people in the isolated ward developed mental trauma and feelings of insecurity. In this chapter, the authors study 90 features of collected data from 207 concerned people during what they faced in lockdown period using machine learning about the insecurity to obtain key features. They have chosen seven ML algorithms like logistic regression, naïve bayes, stochastic gradient descent J48, multi-layer perceptron, random forest, and random tree. These algorithms are used on the features to identify appropriate features on insecurity. Data splitting with 10-fold cross validation reduced 90 features into seven features by comparative analysis. In 66-34 split, 50-50 split, 80-20 split, LR and NB achieved above 90% accuracy with these seven features. 80-20 and in 10-fold we get approximately 100% accuracy in J48, MLP, RF, and RT algorithms. Three feature selection techniques, Information Gain, ReliefF and OneR, were used for ranking features based on model performance.