Analysis of Prognostic Factors Influencing Survival and Recurrence in Breast Cancer: A Hybrid Machine Learning Approach

Abstract Purpose: The present study sought to identify prognostic factors for breast cancer survival and recurrence using a machine learning approach and electronic medical record data.Methods: We used a machine learning technique called feature selection to identify factors influencing breast cancer prognosis, and factors affecting survival and recurrence in a Cox regression model. Results: History of relapse, type of surgery, diagnostic route, SEER stage, and hormone therapy all affected breast cancer survival. Recurrence of breast cancer was affected by age, history of diabetes, breast reconstruction, pain, breast lumps, nipple discharge, and the presence of other symptoms. According to the survival analysis based on feature selection, patients with diabetes had a significantly higher risk of early recurrence of breast cancer (hazard ratio, 4.8; 95% confidence interval, 2.04–11.2, p < 0.05). Conclusions: The present study identified several factors associated with breast cancer prognosis. While survival was affected by the diagnostic route, recurrence was primarily influenced by breast cancer symptoms and other underlying health conditions. A more accurate and standardized model considering time-to-event data could be developed in the future to evaluate prognostic factors and predict prognoses, and for clinical validation