A memetic algorithm based on interval estimation for fuzzy c-means clustering of incomplete data

Clustering as one of the most widely used techniques in data mining is only valid for complete data, while in real applications, many datasets suffer from incompleteness. In this paper a memetic algorithm for incomplete data fuzzy clustering based on the missing values interval estimation is presented. The proposed algorithm combines the genetic algorithm and an adaptive version of hill climbing method for imputing the missing attributes then performs fuzzy c-means approach on the refined data sets. Also, for enhancing the robustness of missing attributes representation, a dynamic approach is applied for determining the best number of data vectors for constructing missing values’ representing intervals. The experimental results for several UCI data sets and their comparison with other methods demonstrate the more efficiency of the proposed method in clustering incomplete data.

کلیدواژه ها:

Fuzzy Clustering ، Genetic Algorithm ، Hill Climbing Algorithm ، Nearest-Neighbors Interval