Method for applying multi-criterion fusion to feature selection of high-dimensional small sample data

Abstract

The invention relates to a method for applying multi-criterion fusion to feature selection of high-dimensional small sample data. The method comprises the following steps of 1) initializing a sample data set and performing clustering on the sample data set; 2) performing feature selection on clustered samples by using a Fisher Ratio method and a Relief F method; 3) fusing feature selection results, applying different penalty factors to classes, and training a PSVM classifier by adopting a fused result; 4) performing regression on the sample data set by adopting the trained classifier, removing features with minimum correlativity, and updating the sample data set; and 5) judging whether coding is ended or not, and if yes, ending the iteration process, or otherwise, repeating the steps 2-4 until the feature selection is realized. According to the method for applying the multi-criterion fusion to the feature selection of the high-dimensional small sample data, in the field of the feature selection of the high-dimensional small sample data, the feature selection speed is remarkably increased, the feature selection efficiency is remarkably improved, and the stability of a feature selection result is greatly enhanced.

Description

technical field [0001] The invention relates to a special selection method, in particular to a method for applying multi-criteria fusion to feature selection of high-dimensional small-sample data. Background technique [0002] Feature selection is one of the core issues in the field of pattern recognition, and its research has attracted the attention of scholars all over the world. The feature selection algorithm achieves the purpose of reducing the feature dimension by selecting effective features reasonably, which can not only eliminate information redundancy, improve classification efficiency, and speed up calculation, but also reduce the complexity and classification error rate of classifiers. At present, feature selection methods have been widely used in image retrieval, text classification and gene analysis. However, the selection results of most current feature selection algorithms are very sensitive to changes in the training set, that is, the algorithm stability is...

AI Technical Summary

Problems solved by technology

However, the selection results of most current feature selection algorithms are very sensitive to changes in the training set, that is, the algorithm is less stable.

Images