Modelling Covid-19 Pandemic in Kenya

Abstract

Due to increased COVID-19 infections and mortality in Kenya, there has been a scarcity of resources like hospital facilities, quarantine centres and personal protective equipment (PPE) for the medical personnel. Therefore, effective planning was required by the Kenyan Government to ensure resources are available to combat the rising COVID-19 cases. This study developed the Autoregressive Integrated Moving Average (ARIMA) and the Holt-Winters models to predict the COVID-19 infections and mortality rates in Kenya. The Quantitative discrete data from the Kenya Ministry of health was used. The data covered 8 months’ period from 23rd August 2020 to 23rd April 2021. The analysis entailed descriptive statistics and the time series prediction technique (ARIMA and Holt-Winters). The COVID-19 infections and mortality data were subjected to the time series prediction models to obtain the accuracy measures for model comparison. The study employed information criteria for model selection. The Simple linear regression was conducted on the COVID-19 data to determine a linear relationship between the COVID-19 infections recorded daily and the samples tested. The regression coefficients were obtained and they were statistically significant at 5% significance level. The R2 of the model was 0.72 implying that 72% of the variation in the response variable is explained by the explanatory variable. The data were fitted to ARIMA and Holt-Winters models using R statistical software (version 4.1.0). The Final model with the lowest Akaike Information Criterion (AIC) was selected. The root mean square error was applied for model comparison and the ARIMA(2,1,3) and the ARIMA(0,1,1) were found to be the best models for predicting COVID-19 infections and mortality with minimum errors. The COVID-19 prediction was done at a 95% significance level using the selected ARIMA models. From the prediction plots, the infections and mortality cases were observed to increase significantly. Therefore, the study suggested that the ARIMA was an effective prediction model for the COVID-19 infections and mortality than the Holt-Winters model. The study recommended that Kenyans should observe the World Health Organization's guidelines to help reduce the infectivity and mortality rate. The government should provide protective equipment for the medical personnel to curb the surging COVID-19 cases. The study also recommended that the ARIMA model be applied for short-term prediction for further studies.