Design of classification model on high-dimensional imbalance data of motor bearing fault

Abstract

According to the characteristics of high-dimensional imbalance distribution of motor bearing fault data, a design scheme of classification model is proposed for the high-dimensional data reduction problem in the classification algorithm. For details: Combining standard particle swarm optimization algorithm and random forest algorithm, a new high-dimensional data reduction algorithm is proposed. Aiming at the imbalance problem of data categories in the classification algorithm, we proposes to use machine learning under the sum of squares of dynamic deviations criterion to divide the minority sample data set into mixed regions, high-purity minority sample regions and outlier regions, and then use smote algorithm to complete the data equalization processing, so as to make the sample data equalization processing more reasonable, Focusing on the task of motor bearing fault classification, a design scheme of using standard particle swarm optimization algorithm to improve the least squares support vector machine model is proposed.