The Comparisons of Prediction Models on Leukemia Incidence and Mortality Age-standardized Rate in Children and Adolescents

In 2022, leukemia is projected to be the leading cancer case globally among children and adolescents, as indicated by the age-standardized rate (ASR) incidence and mortality index. This study aims to identify and predict the ASR incidence and mortality of leukemia using statistical and machine learning approaches, including Generalized Linear Model (GLM), Regression Tree, Random Forest, and Extreme Gradient Boosting (XGBoost). Evaluation metrics used were Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Symmetrical Mean Absolute Percentage Error (SMAPE), and Mean Relative Absolute Error (MRAE). GLM and XGBoost emerged as the best-performing models for ASR incidence, achieving the lowest SMAPE (36.273%) and MRAE (35.870%) scores. For ASR mortality, Random Forest was the top performer with the lowest SMAPE (37.733%) and MRAE (28.395%) scores. Further analysis using the Shapley Additive exPlanations (SHAP) method was conducted to determine the impact of each factor on the models. However, the analysis showed unsatisfactory outcomes due to missing values and the limited number of variables.