Pricevision: the pioneering machine learning framework for the Malaysian Shop Price Index (MSPI)

The dynamics of shop property prices are critical to Malaysia's commercial real estate performance. However, no comprehensive index currently exists to measure this sector. This research introduces the Malaysian Shop Price Index (MSPI), a pioneering data-driven framework developed in collaboration with the Valuation and Property Services Department (JPPH). The study transitionally shifts from traditional Hedonic models, which face linear constraints, to advanced Machine Learning (ML) analytics. Using a robust dataset of 20,702 transactions (2013–2024) from Kuala Lumpur and Selangor, the research identified 22 key physical, locational, and legal attributes influencing value. After rigorous data cleaning and Z-score outlier removal, 18,199 valid cases were analyzed. Variance Inflation Factor (VIF) analysis ensured the removal of multicollinearity, maintaining a strict threshold (VIF < 5). Six ML algorithms which are Linear Regression, Decision Tree, Random Forest, Support Vector Regressor (SVR), XGBoost, and Multi-Layer Perceptron (MLP) were evaluated. Results revealed that Random Forest and XGBoost achieved near-perfect accuracy (R2 =0.997), significantly outperforming legacy Laspeyres Hedonic models. Launched in February 2026, the MSPI serves as Malaysia's first intelligent indicator for commercial property trends. By integrating ML into national property data, this innovation enhances market transparency and provides a superior tool for evidence-based policy and valuation decisions.