INTEGRASI MACHINE LEARNING DENGAN GEOGRAPHICALLY WEIGHTED REGRESSION UNTUK ANALISIS SPASIAL INDEKS PEMBANGUNAN MANUSIA

Abstract

The success of national development is not solely determined by economic growth but also by the quality of human development, which is measured by the Human Development Index (HDI). Although Indonesia's HDI has improved nationally, disparities between regions, particularly in western Indonesiaremain a significant challenge. To address this, a comprehensive analysis is needed to understand the spatial distribution of HDI and its influencing factors. This study proposes an integrated approach that combines Machine Learning techniques, namely K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) with Geographically Weighted Regression (GWR) using adaptive Gaussian weighting to classify and analyze HDI spatially. The KNN algorithm achieved the highest classification accuracy of 94.23%, slightly outperforming SVM at 92.30%. The GWR analysis successfully grouped the study area into 16 spatial clusters with distinct dominant factors, such as population size (JP), human literacy development index (IPLM), per capita expenditure (PPK), open unemployment rate (TPT), and number of districts (JK). For example, the GWR model in South Jakarta identified significant local influences from these variables, with a specific model equation indicating strong spatial dependency. The model's coefficient of determination (R²) reached 84.17%, indicating high explanatory power. These findings demonstrate that the integration of Machine Learning and GWR not only enhances classification accuracy but also provides deeper spatial insights, offering a robust basis for data-driven policy planning aimed at reducing regional disparities in HDI.