Cost-Sensitive Clustering for Uncertain Data Based on Genetic Algorithm

Abstract

The existing clustering methods for uncertain data don`t consider the cost factor, so proposes a method on cost-sensitive clustering for uncertain data based on genetic algorithm (CSUDC). First, give the cost-sensitive learning for uncertain data. Use the interval to dispose continuous and discrete attribute of uncertain data, so the traditional clustering method can cope with uncertain data. Second, a cost-sensitive clustering method for uncertain data is presented. Adopt the real encoding for the clustering data in genetic algorithm. The optimal cluster centers are searched by the selection, the crossover and mutation. The experimental results show, compared to the rest of the several common clustering method for uncertain data, CSUDC has higher accuracy of classification, and takes the total low cost in the clustering process.