POISSON MODEL WITH GAUSSIAN RATE: A STUDY ON PREDICTING COVID-19 CASES IN INDONESIA

• Main Author: Angelica Tjahjono, Fiona
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/84159
• Institution: Institut Teknologi Bandung

"How many positive cases are there today?" and "When will this pandemic end?" were the two frequently asked quesetions during the COVID-19 pandemic—questions that could arise again at any moment. Therefore, with the advancement of modeling tools and the accessibility of data, building predictive models is crucial to face similar and uncertain challenges in the (near) future. This study focuses on developing a predictive model for the daily new COVID-19 cases in Indonesia using Gaussian process with the data ranging from March 2020 to June 2024. The main challenge here is modeling the count data of cases, which requires modification to the Gaussian process. A comparison of used covariance (kernel) functions is also considered to identify the best kernel parameters in explaining counting data. Evaluation results using MAE, RMSE, R-squared, and coverage probability metrics indicate that the initial model performs poorly. However, by using a combination of RBF and periodic kernels, the developed model can adequately capture the count data patterns (up and down trends), especially when kernel parameters are generated from the ACF lag of the data.