Study on Classification Models Utilizing the Multilayer Perceptron Feature Selection and Ensemble Voting Classifier

The existing classification prediction models increase in training time and cost with the increase in the number of features, significantly increase in computational complexity, and irrelevant or redundant features interfere with the model's learning process and affect the accuracy of classification prediction results. In this study, a model combining the multilayer perceptron (MLP) feature selection with an ensemble voting classifier (MRS) is proposed. The MLP feature selection method demonstrates superiority in screening important features, which can effectively extract key information from the original data. Then, the classification prediction performance of the model is further improved by feeding the filtered features into an ensemble voting classifier that incorporates Random Forest (RF) and Support Vector Machine (SVM). The model enhances the processing efficiency of classification problems in the field of data mining, especially in the context of confronting multi-featured and high-dimensional data, and more accurately performs classification prediction. Moreover, the accuracy of the model exceeds that of many other models is proved by the experiments, demonstrating its great potential in solving the classification and prediction problems in the field of data mining.