PARAMETER ESTIMATION OF GSTAR MODEL USING BAYESIAN APPROACH FOR PREDICTING THE RISK OF AIR POLLUTANT

• Main Author: Nurzakiah, Azmi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/65336
• Institution: Institut Teknologi Bandung

Air pollution is one of the problems that is quite a concern in big cities. Especially for the type of air pollutant ????????2.5, the levels often exceed the reasonable limits set by the World Health Organization. This can result in losses, especially in terms of regional morbidity and mortality rates. As a result of decreasing morbidity and mortality rates, community productivity will decrease. To overcome this, it is necessary to anticipate providing the best facilities for sufferers caused by air pollutants. Thus, there is also a role for insurance companies to offer guaranteed products for the public. So that the company does not suffer losses, it is necessary to predict the risks that may occur in the future. So the premium given is quite accurate. This study pays attention to the air pollutant levels ????????2.5 at 25 stations in Seoul City for the 2017–2019 period and applies Generalized Space-Time Autoregressive (GSTAR) modeling with a Bayesian approach. Usually, the estimation is carried out using the least-squares method. The Bayesian approach is carried out as a development in the GSTAR model to obtain a better estimate. Modeling with this Bayesian approach can use the Markov-Chain Monte Carlo numerical method with the Gibbs sampling algorithm. Based on the simulation, parameter estimation using Bayesian approach is better than least-squares method. Furthermore, on the ????????2.5 air pollutant content data, the best model for data with outliers is GSTAR25(1;2) with a binary weight matrix. Meanwhile, the outlier-free data is GSTAR25(1;2) with an inverse distance weight matrix.