Spatial model of breast cancer susceptibility using machine learning in Surabaya, Indonesia

Surabaya City is one of the urban areas in Indonesia with a high incidence of breast cancer. Multiple factors, including genetic predisposition, lifestyle patterns, and environmental exposures, influence this condition. This study aims to model breast cancer susceptibility using machine learning techniques by comparing four algorithms: Support Vector Machine (SVM), Random Forest (RF), Boosted Regression Tree (BRT), and an ensemble approach. The analysis identifies key factors contributing to breast cancer susceptibility and examines their relationships with the observed variables. Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC), True Skill Statistic (TSS), Correlation Coefficient (COR), and deviance. The results demonstrate that the RF model outperformed the individual models, achieving an AUC of 0.93, TSS of 0.70, COR of 0.77, and deviance of 0.53. The three models were subsequently integrated into an ensemble framework, which further improved predictive performance, yielding an AUC of 0.95, TSS of 0.76, COR of 0.76, and deviance of 0.49. Approximately 31.50% of the study area was classified as highly susceptible to breast cancer. Among the twelve identified contributing factors, population density, the Normalized Difference Built-up Index (NDBI), and carbon monoxide (CO) levels were the most influential predictors of breast cancer susceptibility in Surabaya. These findings provide scientific evidence to support the Surabaya municipal government in developing early detection strategies and promoting sustainable urban environmental management.