Fundamentals and Methods of Machine and Deep Learning

4Analysis of M-SEIR and LSTM Models for the Prediction of COVID-19 Using RMSLE

Archith S., Yukta C., Archana H.R.* and Surendra H.H.

Department of Electronics and Communication Engineering, BMS College of Engineering, Bangalore, India

Abstract

During an epidemic period, it is important to perform the time forecasting analysis to track the growth of pandemic and plan accordingly to overcome the situation. The paper aims at performing the time forecasting of coronavirus disease 2019 (COVID-19) with respect to confirmed, recovered, and death cases of Karnataka, India. The modified mathematical epidemiological model used here are Susceptible - Exposed - Infectious - Recovered (SEIR) and recurrent neural network such as long short-term memory (LSTM) for analysis and comparison of the simulated output. To train, test, and optimize the model, the official data from Health and Family Welfare Department - Government of Karnataka is used. The evaluation of the model is conducted based on root mean square logarithmic error (RMSLE).

Keywords: COVID-19, modified SEIR, neural networks, LSTM, RMSLE

4.1 Introduction

Severe acute respiratory syndrome corona virus 2 (SARS-CoV-2), also known as COVID-19 virus, was first found in Hubei province of China in November 2019. Virus was spread to other parts of world to almost all the countries. In India, the first case was reported during January from the state of Kerala. Around 20 lakh individuals are infected by the COVID-19 virus in India and ...