Disease outbreak prediction from time series data using machine learning and deep learning approaches

Abstract

Time series analysis is an important way for understanding complex systems and making informed decisions. Tracking and predicting disease outbreaks all over the world is a complex problem, but time series analysis can be helpful for solving this issue. There are many ways to perform time series analysis like Descriptive analysis, Visual analysis, Frequency domain analysis, machine learning based analysis and many more. In this project, machine learning regression models like Linear regression, Random Forest and deep learning model LSTM(Long short-term memory) is used to forecast time series based disease datasets. Three different datasets(COVID-19 in Kerala, Hungarian chickenpox and Adenovirus in Bay Area) are considered here. From this study, the most accurate model for Covid 19 in Kerala dataset is founded to be the Multiple linear regression model with the RMSE value 2910.811. The most accurate model for Hungarian chickenpox dataset is Random forest(RMSE value 24.799). The most accurate model for Adenovirus in Bay Area dataset is Multiple linear regression (RMSE value 3.1).