Breast cancer recurrence time prediction based on the MOPSO-RF model

Background: Cancer is a complex disease where malignant tumors have high clinical incidence and mortality rates. A significant risk of postoperative recurrence also exists. This work was aimed at diagnosing the time to postoperative cancer recurrence which could help the patients. The post-operative breast cancer recovery data from Beijing Stre Cancer Data Analysis Joint Laboratory was utilized. The study was conducted to determine the weighting of adaptive symptoms regarding the time to breast cancer recurrence by selecting six indicators, i.e., immune, tumour, microenvironmental, psychological, nutritional and aerobic exercise and progressive work, involved in the mechanism of breast cancer recovery. Methods: A multi-objective particle swarm optimised random forest (MOPSO-RF) model for the adjuvant cancer diagnosis was introduced to predict the time to breast cancer recurrence. The random forest model was optimised to verify its accuracy. The multi-objective optimisation combined each indicator with time to the breast cancer recurrence to get an objective function for finding the optimal Pareto solution. Results: The results exhibited that the constructed model was effective in predicting the time to breast cancer recurrence and thus provided early warning indications in this regard. Conclusions: The model achieved 92.17% forecast accuracy compared to the Gradient Boosted Tree (GBDT), Support Vector Machine (SVM), Linear/Logistic Regression (LR), XGBoost and Random Forest (RF) algorithm models.