The feature extracting method via partitioning the feature subspace and data fusion
Hao Huang, Famao Mei, Qinqin Wu, Yuanyi Bao, Hengshan Zhang · 2022
The redundant information contained in feature can be reduced and the accuracy of data analysis is improved via extracting the features from the data set. The existing methods to extract the feature ignoring the information contained in the data vector of feature. In this paper, the similarity between data features is firstly calculated via multiple methods to form the similarity vector of feature. Then the adaptive weighted clustering ensemble is proposed to cluster the similarity vector of feature to partitioning the feature subspaces. Secondly, utilizing the characteristics of the data vector of feature, the weights of the features in the subspace are calculated, and then the effective features are extracted using the linear weighted method. In the experiments, the result shows that the proposed method can significantly improve the accuracy of the data analysis.