PREDICTION OF COMMUNITY TRANSMISSION LEVEL OF COVID-19 USING MACHINE LEARNING ALGORITHMS BASED ON THE CDC SOCIAL VULNERABILITY INDEX

Abstract

Response to hazardous events is crucial in every community, whether natural or anthropogenic disasters. Social Vulnerability Index (SVI) helps people who need support. Social vulnerability refers to the number of adverse effects of external stress, including natural causes or disease outbreaks like the Coronavirus Disease 19 (COVID-19) pandemic on human health. The SVI dataset possesses California state of the US, subdivisions of counties of 15 features into four groups as related themes (i.e., socioeconomic status; household composition and disability; minority status and language; and housing type and transportation). In addition to the SVI dataset, the recent COVID-19 data tracker for each county posted by the Centers for Disease Control and Prevention (CDC) shows the new cases per 100,000 persons in the last seven days. The transmission values are low, moderate, substantial, and high. The impact of SVI on COVID-19 attracts the attention of researchers to find the relationships between SVI and COVID-19 incidence. This paper aims to incorporate SVI data and the incidence in the urban and rural areas of the United States using eight machine learning algorithms for COVID-19 transmission level classification. The experimental results show the proper prediction based on the community transmission level of COVID-19 by considering the features of SVI. Among all used machine learning methods, Random Forest achieved the best performance based on the percentage of various performance metrics accuracy and F1-score.