Penerapan Metode Geographically Weighted Random Forest Regression pada Kasus Tuberkulosis di Provinsi Sumatera Utara

Abstract

Geographically Weighted Random Forest Regression (GWRFR) is an improved method of Random Forest (RF) that considers spatial effects. Spatial effects refer to variations in the relationship between variables across different geographical locations. This is relevant in cases such as tuberculosis prevalence, where differences in social, economic, and environmental characteristics between regions affect infection rates. Tuberculosis (TB) remains a serious health problem in Indonesia. According to the Global Tuberculosis Report 2022, North Sumatra Province ranks fourth as the area with the highest number of TB cases in Indonesia, after West Java, East Java, and Central Java, with a total of 83,949 TB cases. This research aims to analyze the spatial distribution pattern of TB cases and identify the factors that contribute to the high number of cases by applying the Geographically Weighted Random Forest Regression (GWRFR) method using an adaptive kernel weighting function. The analysis results indicate that there are eight groups of variables that have a significant impact on TB cases in North Sumatra Province. The grouping of these variables is based on the three most dominant factors in each district/city. From the overall results, the number of health facilities is the most dominant factor affecting the distribution of TB cases across all districts/cities analyzed. This indicates that the number of health facilities should be considered in the management of TB cases in each district/city in North Sumatra Province.