A Clustering-Based Optimization of Restaurant Locations in Malaysian Cities

Abstract

Restaurant location planning in Malaysia's urban centres is becoming increasingly complex due to ethnic diversity, economic disparity, and shifting consumer behaviour. Traditional methods, based on intuition or static foot traffic data, often fall short in addressing this multi-layered reality. This study applies a clustering-based approach to optimize restaurant placements using geospatial, demographic, service, and pricing data across urban districts in Malaysia. Using K-means clustering, the analysis segments over 12,000 restaurant entries extracted from Google Maps and merged with 2024 population projections from the Department of Statistics Malaysia into three meaningful market types. Cluster 1 captures premium urban zones with higher price points and strong customer reviews, typical of affluent districts like Kuala Lumpur and Petaling. Cluster 2 represents stable, budget-sensitive markets with consistent service patterns and affordability-focused offerings. Cluster 3, a hybrid category found in suburban districts such as Kuantan and Seberang Perai, balances moderate pricing with high consumer satisfaction, reflecting evolving middle-class expectations. The model's robustness was validated through Silhouette score evaluation, yielding an average Silhouette score of approximately 0.38, indicating moderate cluster cohesion. PCA clearly visualized cluster separation, while spatial distribution mapping confirmed distinct geographic patterns corresponding to each cluster. Results reveal distinct demand structures and pricing trends across clusters, offering strategic insights for restaurateurs and urban planners. This study not only enhances data-driven decision-making but also demonstrates the value of integrating unsupervised learning methods into commercial geography within Southeast Asian contexts.