GA-SA-Blending Fusion Prediction Model Based on PMRCL Fusion Feature Selection

This paper proposes a GA-SA-blending fusion prediction model based on genetic algorithm (GA) and adaptive selection (SA) to improve the prediction accuracy and stability of the machine learning (ML) model. More specifically, we use diverse feature selection methods, encompassing Pearson correlation coefficient (PCC), maximal information coefficient (MIC), random forest (RF), gradient boosting algorithm (CatBoost), logistic regression (LR), and out-of-bag (OOB) data error estimation method. Our proposed PMRCL fusion feature selection method effectively discards redundant features while optimizing data quality. Meanwhile, the introduction of a genetic algorithm further improves the accuracy and stability of the model in the adaptive selection of the optimal base learner and parameters. The experimental results indicate that the proposed GA-SA-blending model presents significant advantages in prediction accuracy, generalization ability, and stability compared with eight mainstream machine learning (ML) models such as traditional support vector machine (SVM) and XGBoost. Briefly, the proposed model reveals the potential of ensemble learning in feature selection and base learner optimization.