A hybrid optimal clustering algorithm for datasets with missing attributes

In order to improve the accuracy of fuzzy clustering of datasets with missing attributes, a hybrid clustering algorithm is proposed in this paper. The algorithm determines the estimation intervals of missing attributes by the Nearest Neighbor method and the ranges of cluster centers by seeking the maximum and minimum of attributes, which constitute the constraint space together. Under the space, the estimated values of the missing attributes and the cluster centers are obtained simultaneously by ant colony algorithm. Then data are clustered by the membership function of Fuzzy C-means algorithm. Two standard datasets from UCI database are used for the simulation experiment as the samples, in which the effectiveness of the algorithm is proved through the comparison with the four methods proposed by Bezdek.