Unsupervised feature selection algorithm based on spectral clustering and analysis of variance

摘要

Now many data mining tasks have to deal with high dimension problems. Feature selection is put forward to solve such problems, the original data set is handled with feature selection, which can facilitate clustering, classification and other process of data mining by the relevant feature subset. In reality, the label of data set is difficult to obtain or cost relatively expensive in many cases, so unsupervised feature selection algorithm were proposed. This paper mainly focus on the unsupervised feature selection algorithm based on spectral clustering of improvements, the analysis of variance is introduced to the eigenvectors space by Spectrum mapping. Experiments show that the improved strategy can effectively increase the effect of unsupervised feature selection algorithm based on spectral clustering.