Prediction of Potential Food Vulnerability and Food Security in Pekanbaru City Using the Geographically Weighted Regression Method

Abstract

Food vulnerability reflects a region’s inability to adequately meet community food needs in terms of availability, accessibility, and utilization. This study aims to analyze the factors influencing food vulnerability and to compare the performance of the global regression model and the Geographically Weighted Regression (GWR) model in modeling food vulnerability in Pekanbaru City, Indonesia. This study employs secondary data obtained from the 2023 Pekanbaru Food Security and Vulnerability Report, covering 83 urban villages as the unit of analysis. The independent variables include the Priority of Infrastructure Ratio, Priority of Poor Population Ratio, Priority of Road Access, Priority of Households without Access to Clean Water, and Priority of Population per Health Worker Ratio, while Composite Priority is used as the dependent variable. The analysis was conducted using multiple linear regression as the global model and GWR as the local model to capture spatial heterogeneity. The results of the ANOVA analysis indicate that the global regression model produces a residual sum of squares of 3,208,363.416, suggesting that a considerable proportion of food vulnerability variation remains unexplained. The GWR model demonstrates superior performance, with an R² value of 0.820164 and an adjusted R² value of 0.773799, both higher than those of the global regression model. Additionally, the GWR model produces lower Akaike Information Criterion (AIC) and corrected AIC (AICc) values of 556.080285, indicating a better balance between model accuracy and complexity. These findings confirm that spatial heterogeneity significantly influences food vulnerability patterns. Therefore, spatially targeted and location-specific policy interventions are required to effectively reduce food vulnerability across urban villages in Pekanbaru City