Predicting shopping mall tenant density and occupancy through magnet factor analysis using decision tree models

Purpose This study develops a data-driven framework to analyze the occupancy performance of 100 Thai shopping centers, integrating 18 locational, economic, and physical variables to address gaps in retail real estate research. Design/methodology/approach The research applies the Decision Tree algorithm to classify occupancy into ‘Good’ and ‘Bad’ categories, comparing its performance against Gradient Boosted Trees and Naïve Bayes models for transparent rule interpretation. Findings Decision Tree achieved 75.00% accuracy. Renovation emerged as the primary reactive determinant. For non-renovated malls, strong food/retail facilities and foreign customer focus yielded the highest occupancy reliability. Research limitations/implications The study focuses exclusively on 100 locations in Thailand, limiting direct cross-national generalizability. Future research should incorporate longitudinal tracking and broader regional datasets. Practical implications Practitioners gain a transparent, actionable decision-rule framework to assess site viability, optimize tenant and facility mixes, and mitigate investment risks in retail properties. Social implications Optimized mall planning enhances community retail accessibility, supports local merchant integration, and promotes sustainable urban commercial spaces. Originality/value This paper offers an interpretable machine-learning classification model for retail occupancy without needing post-hoc explainability tools, uncovering reactive renovation dynamics.