Multi-view Fuzzy Clustering Algorithm Based on Non-Negative Matrix Factorization and Partition Adaptive Fusion

Nonnegative matrix decomposition (NMF), as a new method of matrix decomposition and feature extraction, is an effective method of clustering by linearly separating data in high-dimensional big data processing and pattern recognition. Aiming at the challenge of how to better realize the cooperation among perspectives in Multi-view clustering task, a new perspective fusion strategy is proposed. This strategy firstly sets a partition for each perspective, and then obtains a fusion weight matrix through adaptive learning to carry out adaptive fusion for the partition of each perspective, and at the same time integrates the feature extraction process into the clustering task, and finally obtains the global partition result by using the perspective integration method. The above strategies are applied to the classical FCM fuzzy clustering framework, and Alternating Direction Method of Multipliers (ADMM) optimization model is adopted to solve the problem. Experimental results on simulated data sets and real data sets have shown that the proposed algorithm has better adaptability and better clustering performance than several correlation clustering algorithms when dealing with multi-perspective clustering tasks.