XGBoost-based model for predicting five-year survival in gastric cancer using clinical indicators

Abstract

Accurate diagnosis and prognostic assessment of gastric cancer are critical for improving patient outcomes. The application of advanced machine learning methods, particularly the XGBoost algorithm, offers a promising approach for enhancing prognostic evaluations. In this study, data from 2,270 patients with gastric cancer were analysed to develop a predictive model for prognosis using the XGBoost algorithm and 20 key clinical features. Comprehensive data collection, preprocessing, and feature selection were conducted to ensure robust model construction and validation. The model demonstrated strong predictive performance in the test cohort, achieving an area under the curve (AUC) of 0.855, and it effectively differentiated patients at high-risk from those at low-risk. Feature importance analysis revealed that pTNM stage and CA125 level were the most influential prognostic factors. This study successfully implemented a machine learning-based model integrating the XGBoost algorithm and critical clinical indicators to predict the five-year survival rate of patients with gastric cancer. The findings highlight the potential of such approaches in supporting personalised treatment strategies and advancing cancer prognosis assessment methodologies.