Figures from this paperAbstractFigures from this paperAbstractFigures from this paperAbstractFigures from this paper

Abstract

This report aims to explore, design, and propose new data mining methods based on structural information with the maximum interval idea, from the construction of non-parallel hyperplane SVM optimization models, extraction and expression of sample structure information, and fast solving methods for learners. So far, domestic and foreign scholars have conducted some preliminary research on non-parallel hyperplane SVM and achieved certain research results. Aiming at the feature selection problem of data classification in machine learning, another method of Twin Support Vector Machine (TWSVM), LFTWSVM, is proposed. Firstly, after solving the TWSVM optimization problem, two weight vectors will be obtained. These two weight vectors will be normalized first, and then the absolute values of the processed two weight vectors will be added to obtain a total weight vector. Finally, the total weight vector will be used for feature selection. Through experiments, the obtained data results are compared with the TWSVM feature selection method, and the LFTWSVM feature selection method has certain advantages.