Multi-Objective Planning Model Based on the Soft Voting Ensemble Learning Algorithm

Ensemble learning is a machine learning approach that utilizes a series of learners for training and integrates their results using certain rules to achieve better learning performance than individual learners. The study employs ensemble learning approach with a Soft Voting ensemble learning algorithm to solve a concrete problem about bank loan. We first analyze and extract features, four major categories comprising a total of 20 features are considered to assess credit risk. Then the Soft Voting ensemble learning algorithm is established, in this paper, Adaboost, Logistic Regression, Random Forest, GBDT, and SVM are utilized as base classifiers. Through experiment, the comprehensive accuracy of the Soft Voting algorithm on the entire dataset exceeds 95%, with a Receiver Operating Characteristic (ROC) curve area under the curve (AUC) of 98%, which indicates high model credibility. Additionally, a multi-objective planning model for loans is established, which maximizes profit while minimizing risk through the credit risk value.