A novel algorithm of data mining to predict future scenarios of COVID-19 pandemic

Abstract

COVID-19, a novel coronavirus, is an ongoing global pandemic that has outbroken recently and spread to almost every part of the world. Several factors of this pandemic are still unknown to the world, which causes uncertainty to prepare a strategic plan to cope with this disease effectively and securing the future. A large number of research is in progress or expected to start shortly on the basis of the publicly available datasets of this deadly pandemic. The data are available in multiple formats that include geospatial data, medical data, demographic data, and time-series data. In this study, we propose a data mining method to classify and forecast the time-series pandemic data in an attempt to predict the expected end of this pandemic in a particular region. Based on the COVID-19 data obtained from several countries around the world, a naïve Bayes classifier is built, which may classify the affected countries into one of the following four categories: critical, unsustainable, sustainable, and closed. The pandemic data collected from online sources are preprocessed, labeled, and classified by using different data mining techniques. A new clustering technique is also proposed to predict the expected end of the pandemic in different countries. A method to preprocess the data before applying the clustering technique is also proposed. The results of naïve Bayes classification and clustering techniques are validated based on accuracy, execution time, and other statistical measures.