Recommended Citation
Xu, Peter, "Machine Learning to Simulate and Predict Infection Rates of Covid-19 Pandemic" (2021). Discovery Undergraduate Interdisciplinary Research Internship. Paper 28.
https://docs.lib.purdue.edu/duri/28
Date of this Version
12-20-2021
Keywords
Machine Learning, Covid-19 Pandemic, Hawkes Process
Abstract
As the covid-19 pandemic brings uncertainty to the medical system, engineering technology comes ahead in an attempt to understand and simulate the impact of it. As the unprecedented pandemic worsens in some states, the medical system encounters severe shortage of medical supplies. To better manage the limited resource nation-wise, the system must foresee the demand in each state and distribute the resources accordingly and prematurely. This project aims to simulate the pandemic and make short-term prediction on daily infections in each state. This forecast would allow the medical system to recognize the states that would require supplies of hospital resource in the near future. Its exact predictions on infection counts would also decide the amount of supply allocation to each state. The project takes historical infection rates and implements Expectation-Maximization algorithm to configure the parameters for the Hawkes Process model – the underlying model for the simulation and prediction. Then, thinning algorithm is used to realize the simulation and to forecast the severity of pandemic. The goal of the project is to make more precise prediction on the infection rates, so the allocation of resource in the system would be optimal in addressing the shortage of supplies across the nation.