A Hybrid LSTM and MLP Scheme for COVID-19 Prediction: A Case Study in Thailand

• Published in: Trends in sciences Vol. 20; no. 10; p. 6884
• Main Authors: Kompunt, Peerapol, Yongjoh, Suparat, Aimtongkham, Phet, Muneesawang, Paisarn, Faksri, Kiatchai, So-In, Chakchai
• Format: Journal Article
• Language: English
• Published: 01.09.2023
• ISSN: 2774-0226; 2774-0226
• DOI: 10.48048/tis.2023.6884

After the COVID-19 epidemic, Thailand was affected in a variety of ways, with the most obvious being the economic downturn and the huge impact on health, including the loss of medical and human resources to combat the epidemic. However, Thailand still lacks analysis and prediction tools required to prepare for future epidemic situations. Therefore, we present development models for predicting the spread of the COVID-19 epidemic. In particular, the application of a long short-term memory (LSTM) and multilayer perceptron (MLP) model was investigated to predict new cases, total cases, new deaths, and total deaths. There are a total of 77 provinces in Thailand. The data used in this trial were obtained from the Department of Disease Control (DDC) of the Thai government. The modeling employed 2 types of data: dynamic (time series) and static. There were 2 phases: 1) the LSTM was used to manipulate time series data and 2) the MLP model was used to manipulate static data. Then, the models were merged for further analysis. We evaluated the performance of the combined model, yielding an accuracy of 99.72 % based on R2 values, higher than the values obtained for state-of-the-art methods. In addition, the prediction results can be further combined with GIS data in each province and displayed via an easy-to-use web application for mapping. HIGHLIGHTS A new architecture that can be used as a tool to predict the COVID-19 epidemic situation is proposed Deep learning is applied to create predictive models with LSTM for time series data and MLPs for static data A tool used for displaying predictions using a web application in the form of a map of each province in Thailand is developed to show the prediction results in detail GRAPHICAL ABSTRACT