Forecasting COVID-19 cases using Machine Learning models

• DOI: 10.1101/2020.07.02.20145474
• Publication Date: 2020-07-05T10:34:35Z
• Authors: Yuan Tian; Ishika Luthra; Xi Zhang
• Keywords: Social distance

Abstract As of April 26, 2020, more than 2,994,958 cases COVID-19 infection have been confirmed globally, raising a challenging public health issue. A predictive model the disease would help allocate medical resources and determine social distancing measures efficiently. In this paper, we gathered case data from Jan 22, 2020 to 14 for 6 countries compare different models’ proficiency in prediction. We assessed performance 3 machine learning models including hidden Markov chain (HMM), hierarchical Bayes model, long-short-term-memory (LSTM) using root-mean-square error (RMSE). The LSTM had consistently smallest prediction rates tracking dynamics incidents 4 countries. contrast, provided most realistic with capability identifying plateau point growth curve.