Implementation of Latent Gaussian Model in the Forecasting Process

Abstract

This research aims to determine the implementation of the Latent Gaussian Model in the forecasting process. This research focuses on developing a forecasting model using the Multivariate Latent Gaussian Model (LGM) approach with shared components. which offers a more accurate representation without the assumption of stationarity and cointegration as it accommodates random components in the model. The forecasting results for the five KPPs are considered to have a very good level of accuracy with MAPE values < 10%. This shows that LGM can achieve reliable forecasting when applied to the real life problems. This condition supports forecasting and can be an effective and targeted benchmark. The Latent Gaussian Model using the Bayesian Approach in parameter estimation can be utilized in forecasting Personal Income Tax Article 25/29. This is supported by the highly accurate MAPE value of 0.01%. The implementation of the developed model is not limited to forecasting Personal Income Tax Article 25/29. but can also be used in various other fields. With its hierarchical structure. the Bayesian approach proves to be an effective method for addressing complex modeling challenges.