Prostate cancer is a prevalent malignant disease among middle-aged and elderly men. Its prevention and detection are significant public health issues. We aimed to construct an interpretable model for predicting death risk in prostate cancer patients. We performed model development using the Cancer Genome Atlas and the Genotype-Tissue Expression databases. In comparison among models, the SVM model has the highest prediction performance among the eight models. The SHAP method, sorted by importance, reveals the top eight predictors of prostate cancer disease. This effective computer-aided approach can facilitate frontline clinicians in the diagnosis and management of patients with prostate cancer.
Keywords: Prostate cancer; Shapley Additive Explanations (SHAP); machine learning; predictive model; support vector machine (SVM).