Multitask possibilistic and fuzzy co-clustering algorithm for clustering data with multisource features

People often encounter two major problems for the practical clustering problems. One is the problem arising from improper extraction of feature sets, such as the weakness of the features and the feature vector usually has the property of high-dimensional and multisource. The other is that the outliers interfere with the clustering results. In this paper, we use the idea of co-clustering to cluster datasets and feature sources at the same time, and use the information which received from the information sharing between tasks to improve the accuracy of clustering tasks through the idea of multitask. And we used the advantage of the typical degree to construct a new parameter selection index to identify the outliers, and to correct each parameter by weakening the influence of the identified outliers on the clustering results. In order to reflect the applicability and robustness of the algorithm, we extend the algorithm to the non-precise dataset and evaluate the algorithm from multiple aspects through experiments. Experiments show that the proposed algorithms not only improve the clustering accuracy, but also greatly reduce the interference of outliers to clustering results.