Skip to Main Content
Presents an unsupervised feature selection method using a fuzzy-genetic approach. The method minimizes a feature evaluation index which incorporates a weighted distance used to rank the importance of the individual features. In addition, a fuzzy membership function is employed to determine the degree of closeness for each pair of patterns which are used in the feature evaluation index. A genetic algorithm is then applied to find an optimal set of weighting coefficients that minimizes the evaluation index. The final weighting coefficients denote the importance of each feature. Several experimental results are given.