SEASONAL GENERALIZED SPACE-TIME AUTOREGRESSIVE MODELING ANDTRANSFER FUNCTION FOR ANALYZING RAINFALL RISK ON RICE PRODUCTIVITYIN WEST JAVA

• Main Author: Afra Nailah Adma, Fathimah
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82374
• Institution: Institut Teknologi Bandung

Models such as Vector Autoregressive (VAR) and Generalized Space-Time Autoregressive (GSTAR) are commonly used in space-time data analysis. However, these models are unsuitable for analyzing space-time data with seasonal patterns. Consequently, this study develops a Seasonal GSTAR (SGSTAR) model to handle space-time data with seasonal patterns for a variable across multiple locations. Specifically, the spatial order is restricted to one. The optimal model identified is SGSTAR(1;1)(3;1,1,1)4, which is then employed to forecast monthly rice productivity for the next 12 months in six districts in West Java: West Bandung, Bandung, Ciamis, Cirebon, Garut, and Tasikmalaya. Subsequently, the impact of rainfall on rice productivity in these six districts is analyzed. A Single-Input Transfer Function model is utilized to predict rice productivity (output series) using rainfall at each location’s input series. The risk associated with these rice productivity forecasts and historical data is analyzed using an Early Warning System.