The Study And Application Of New Clustering Algorithms In Image Processing And Text Clustering

Patten recognition is one of important tasks of artificial intelligence research, which has been extensively studied in the past tens of years. With the rapidly development of data mining and data analysis technologies, clustering analysis methods have been generally utilized in various fields such as pattern recognition, image processing, computing vision, and so on. The existed various clustering methods have been applied in different applications, with their distinctive advantages respectively. Nowadays, how to reduce the sensitivity of the algorithms to noise while obtain the optimal robust cluster approaches, and properly explain the clustering procedure are some top issues that many scholars are addressing.Motivated by the above challenges, several issues are addressed in this study, which mainly involves the following three parts.In Chapter 2, a generalized fuzzy c-means clustering algorithm with improved fuzzy partitions is proposed to overcome the shortcoming of the existing IFP-FCM algorithm. By introducing the modified fuzzy partition constraints, a new objective function is constructed and furthermore the generalized FCM algorithm with improved fuzzy partitions (GIFP-FCM) is derived. Meanwhile, from the viewpoints of Voronoi distance and competitive learning, the robustness and convergence of the proposed algorithm are analyzed. More importantly, the classical FCM and the IFP-FCM can be taken as two special cases of the proposed algorithm. Several experimental results including its application to noisy image texture segmentation are presented here to indicate its average advantage over algorithm FCM and IFP-FCM in clustering and robust capability.In Chapter 3, one of the important characteristics of the text clustering datasets may perhaps be that a cluster center in the dataset has a distinctively different direction from all other cluster centers. For such situations, directional information should be well-incorporated into the corresponding clustering approach. A new robust fuzzy directional similarity clustering algorithm (RFDSC) is proposed. By introducing the fuzzy partition constraints, a new objective function is constructed and furthermore RFDSC is derived. Meanwhile, from the viewpoints of competitive learning, the robustness and convergence of the proposed algorithm are analyzed. Texts clustering experimental results demonstrate the effectiveness of RFDSC.In Chapter 4, a new robust Improved Fuzzy Partitions for K-Plane Clustering (IFP-KPC) algorithm is proposed. The proposed algorithm can reduce the sensitivity of the k-plane clustering algorithm to noises in real datasets. Also the distances to the Voronoi cell are used to give a reasonable explanation for the robustness of IFP-KPC. Experimental results demonstrate the effectiveness of IFP-KPC.