Background: Cancer remains a leading cause of morbidity and mortality worldwide, with significant variations in incidence, mortality, and survival rates across regions. This study leverages Machine Learning (ML) to analyze global and Iran-specific cancer data, aiming to improve predictive accuracy for cancer mortality. Methods: Using datasets from Global Cancer Observatory (GLOBOCAN) and the Iran National Cancer Registry (INCR), we evaluate the performance of ML models, including XGBoost, Random Forest, and Support Vector Machines, in predicting cancer outcomes. Results: XGBoost achieved superior performance globally (R2 = 0.83, AUC-ROC = 0.93) compared to Iran-specific data (R2 = 0.79, AUC-ROC = 0.89), highlighting the influence of region-specific risk factors such as Helicobacter pylori infections in Ardabil. Additionally, we explore the application of ML in predicting Second Primary Cancer (SPC) risk, emphasizing the role of radiation dose, age, and genetic mutations as key predictors. Conclusion: This research underscores the potential of ML to inform personalized treatment plans and improve cancer care while addressing challenges such as data imbalances and regional disparities. The findings provide valuable insights for policymakers, researchers, and healthcare providers in developing targeted interventions to reduce the global cancer burden.
