Leveraging Data Mining Strategies for Predictive Modeling of COVID-19 Outbreak and its Socio-economic Impacts: A Simulation for Public Health in Iran

We present a thorough analysis of socio-economic impacts of COVID-19 on public health through data mining strategies including correlation index matrix, auto-regressive integrated moving average, decision trees, heatmaps and statistical performance evaluation. We acquired and filtered data for mortality and outbreak prediction through key features such as total cases, daily new cases, active cases, total deaths, daily new deaths, newly recovered, death rate and recovery rate for 54 days. The socio-economic impacts of the pandemic through quantitative analysis of stock market index, currency inflation, gasoline prices, interest rate, consumer price index and crude oil prices were also investigated. With correlation index matrix and heatmaps, we discovered the nature and intensity of interdependency of these features and developed the regressive estimation model to forecast the values of inter-related features for 10 days. We observed a highest correlation of +0.95 between recovery rate and total infected cases. We also observed an inverse correlation of -0.81 between daily new cases and recovery rate due to unexpected rise in outbreak. Also, the mild but positive index for economic impacts, such as currency inflation, depict the virus’ adverse impact on the fiscal situation. The statistical representation of the developed prediction models through bar charts show outstanding performance when evaluated on the benchmarking merits of mean absolute error, root mean square error, relative error and percentage accuracy.