COVID-19 CASES PREDICTION AFTER VACCINATION USING SEIVR MODEL AND LATIN HYPERCUBE SAMPLING

• Main Author: Isfan Rahadi, Muhammad
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/72097
• Institution: Institut Teknologi Bandung

Since it was first identified, the COVID-19 virus has become a common problem for most countries in the world, including Indonesia. Therefore, predictions from the course of COVID-19 are crucial for policy makers to optimize virus management. Currently, there are various types of models to predict the spread of COVID-19 disease but most models do not adopt vaccination as a calculation factor. The SEIVR model is a compartmental model for direct infectious diseases that calculates the vaccination factor in its prediction process. In this study, a modification of the SEIVR model has been produced to be applicable to data in Indonesia. One form of modification is to add the dead population to the SEIVR model. In addition, the SEIVR model as a deterministic predictive model has a quality weakness. Therefore, this research transforms the SEIVR model from deterministic to probabilistic. In carrying out the thesis research, the method was divided into 5 main stages, namely data collection, data pre-processing, SEIVR model building, model transformation, and evaluation. In each main stage several other methods are used. The regression method was used in determining variable values in the SEIVR model with R Square and RMSE evaluations. Meanwhile, the Latin Hypercube Sampling method was used in the model transformation process to determine the value of the upper and lower bounds of the middle percent. The evaluation results of the research found that modifications to the SEIVR model can be used to predict the number of COVID-19 in Indonesia. In addition, by transforming the deterministic SEIVR model to the SEIVR probabilistic model, the overall model accuracy increased by 8.7% and reduced the average distance of the largest error prediction from the actual data by 0.005.