Track: Modeling and Simulation

Abstract

This paper extends the research on the accuracy of the Long Short Term Memory (LSTM) deep learning model to forecast Covid-19 daily cases in the Philippines with features affecting the transmission of the virus. The study utilized a mixed method; qualitative research to analyze and understand experiences to generate ideas and quantitative research to gather measurable data that will be used for statistical analysis.  The study examined three (3) factors affecting the forecasting of Covid-19. Moreover, the modeling procedure was completed using python 3.0. The research results indicate that the forecasting of Covid-19 daily cases using the LSTM forecasting model has an error of no more than 17% error which depicts as an acceptable accuracy. It also specifies that the model shows more meaningful trends in the two (2) week forecast rather than the two (2) month forecast. The study triangulated the trends of forecast with the help of the focus group discussion with health experts and health administrative personnel. The results specify that 75 percent wear the necessary PPEs when going out and 60 percent always keep physical distance. Clearly, the strictly enforced protocols have been adapted by the Filipinos. Furthermore, with the given percentage error of the model, the significant factors and the optimal hyper parameters incorporated in the model helped in the development of the model. Using the LSTM deep learning model, the study provides a framework that helps understand the behavior of the virus and forecast Covid-19 cases months ahead.

 Keywords: Long Short Term Memory (LSTM), Covid-19, Python, Healthcare, Forecasting