Partial Least Squares Structural Equation Modeling with Prediction-Oriented Segmentation for Analyzing Community Welfare in East Java Province

This study examines the relationships between infrastructure development, community empowerment, government governance, and community welfare across 38 regencies and cities in East Java, Indonesia, using secondary data from 20222024. Partial Least Squares Structural Equation Modeling (PLS-SEM) was used to analyze these relationships, while Partial Least SquaresPrediction-Oriented Segmentation (PLS-POS) explored regional heterogeneity. Infrastructure development was positively and significantly associated with community welfare and government governance, but negatively associated with community empowerment. Community empowerment and government governance were not significantly related to community welfare, and none of the tested indirect effects was significant. PLS-POS identified two exploratory segments: Segment 1 (14 regions; 36.84%) had relatively higher welfare and infrastructure development scores but lower empowerment and governance scores, whereas Segment 2 (24 regions; 63.16%) showed the opposite pattern relative to the overall mean. Most structural relationships showed similar patterns across segments, although the community empowermentwelfare relationship differed in sign. Run 12 achieved the highest objective criterion value among 20 runs, but variation in segment membership across random starts and the absence of formal between-segment significance tests limit the conclusions. These findings should therefore be interpreted as exploratory. Future research should use larger datasets and formal tests, such as Multi-Group Analysis (MGA) or permutation tests, to assess whether segment differences are statistically significant.