CONSTRUCTING MALAYSIAN SHOP PRICE INDEX (MSPI) USING MACHINE LEARNING ANALYTIC METHOD
Sr Dr. Junainah Binti Mohamad (Project Leader)
Professor Sr Dr. Thuraiya Mohd
Assoc. Professor Ts Dr. Suraya Masrom
Dr. Lizawati Abdullah
Najma Binti Azman
Nurul Afiqah Anuar
Dr. Shazmin Shareena Binti Ab. Azis
Intan Faiqah Hamizah Binti Mohd Firazan (Research Assistant)
(Universiti Teknologi MARA)
ABSTRACT
This study develops the Malaysian Shop Price Index (MSPI) using machine learning (ML) techniques to provide an adaptive, data-driven alternative to traditional hedonic models for tracking shop property price movements. The research has three main objectives which are to identify factors influencing shop property prices, to profile the MSPI, and to construct the MSPI using ML algorithms for national-level prediction. Through an extensive review of existing studies, 22 variables were identified across physical, locational, and legal categories, with transaction price (actual and logarithmic) as the dependent variable. Using 18,199 valid transactions from Kuala Lumpur and Selangor (2013–2024), the study conducted correlation and Variance Inflation Factor (VIF) analyses to ensure model accuracy. Six ML algorithms which are Linear Regression, Decision Tree, Random Forest, Support Vector Regressor (SVR), XGBoost, and Multi-Layer Perceptron (MLP) were employed, with Random Forest and XGBoost achieving near-perfect accuracy. This research establishes the MSPI as a reliable, transparent tool for monitoring shop property price trends, enhancing market transparency and supporting evidence-based valuation and policy decisions.
Keyword: Malaysian Shop Price Index (MSPI); Machine Learning; Commercial Property Valuation; Hedonic Price Model; Real Estate Market Analysis