Predicting Early Lease Termination Risk in Jakarta Shopping Malls Using a SMOTE-Enhanced SVM Model for Financial Loss Prevention

The high incidence of early lease termination in shopping malls poses significant challenges to revenue generation, unit utilization, and the operational stability of commercial properties. The limitations of traditional management practices in identifying high-risk tenants early often result in financial losses and suboptimal asset allocation. To address this issue, this study developed a data-driven predictive model designed to identify the likelihood of early lease termination. The approach integrates the Support Vector Machine (SVM) algorithm with the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance within the dataset. The model development followed the CRISP-DM methodology and utilized a historical dataset comprising 795 lease records from a major shopping mall in Jakarta, spanning the years 2015 to 2022. Through systematic data preprocessing, feature selection, and model optimization using grid search and cross-validation, the model achieved excellent classification performance: 93.10% accuracy, 90.50% precision, 96.40% recall, 93.30% F1-score, and 97.30% AUC. The findings demonstrate that the SMOTE–SVM combination consistently outperforms in detecting minority-class cases. A prototype system was also developed, enabling mall managers to predict tenant risk in real-time through an intuitive user interface. The contributions of this research are twofold. First, it presents a novel application of the SMOTE–SVM approach for addressing data imbalance in early lease termination prediction within the Indonesian commercial property sector an area that remains underexplored. Second, the study delivers a practical and deployable prototype system that enables real-time risk assessment for mall management, thereby bridging the gap between predictive modeling and operational decision-making. Overall, the proposed model offers a reliable and scalable predictive solution that can be adapted for risk management in other commercial property contexts, supporting a data-driven and proactive decision-making approach. However, it is important to note that the applicability of the proposed SMOTE–SVM model may face certain challenges when deployed in different commercial property contexts. Variations in tenant characteristics, market dynamics, economic conditions, and data availability across regions could impact model generalizability and performance. Moreover, the reliance on historical lease data assumes consistency in tenant behavior patterns, which may not hold true in rapidly evolving retail environments or for properties with distinct operational models such as coworking spaces or mixed-use developments. These factors should be carefully considered when adapting the model to ensure its validity and effectiveness outside the original study setting.