<![CDATA[One of the models that is utilized in spatio-temporal analysis is known as the Generalized Space-Time Autoregressive (GSTAR). This model incorporates two dimensions, namely the geographical and temporal aspects of the situation. This approach assists in the identification of patterns and correlations between data by taking into account both spatial and temporal elements. From modeling the confidence level of forest fire hotspot cases in Kubu Raya and its surrounds using the GSTAR (1;1) model with three different combinations of grids and special weight matrices, the purpose of this study is to discover which combination of grids and spatial weight matrices is the most effective. The results of diagnostic tests and the degrees of MAPE accuracy are used to determine which model is the most suitable. The data was obtained from the FIRMS-NASA platform, ranging from January 2014 to August 2024. A grid with a dimension of 1.25 x 1.25 degrees and a rook contiguity weight matrix is a combination of grids and spatial weight matrices that meet the white noise assumption, according to the findings of the study. This conclusion is based on the diagnostic test. As a result, the combination of a grid with a size of 1.25 x 1.25 and a rook contiguity weight matrix is the best in this modeling. This combination has a MAPE of 11.797%, which indicates that this model has a good level of accuracy. ]]>
Copyrights © 2025
