Streaming Data Clustering using Incremental Affine Propagation Clustering Approach

Clustering domain is vital part of data mining domain and widely used in different applications. In this project we are focusing on affinity propagation (AP) clustering which is presented recently to overcome many clustering problems in different clustering applications. Many clustering applications are based on static data. AP clustering approach is supporting only static data applications, hence it becomes research problem that how to deal with incremental data using AP. To solve this problem, recently Incremental Affinity Propagation (IAP) is presented to overcome limitations. However IAP is still suffered from streaming data clustering support missing. In this project our main aim is to present extended IAP with support to streaming data clustering. This new approach is called as IAP for Streaming Data Clustering (IAPSDC). First we have to present two IAP clustering methods are presented such as K-medoids (IAPKM) as well as IAP clustering using Nearest Neighbor Assignment (IAPNA). For streaming data with IAP we are using our algorithm for clustering streaming data uses a subroutine called LSEARCH algorithm. The practical work for this project will conducted on real time datasets using Java platform. Though many clustering problems have been successfully using Affinity Propagation clustering, they do not deal with dynamic data. This paper gives an incremental clustering approach for a dynamic data. Firstly we discuss the affinity propagation clustering in an incremental space using K-medoids and nearest neighbour algorithm and then propose an algorithm Incremental Affinity Propagation using Streaming Data Clustering (IAPSDC) using the same approach in streaming data clustering. IAPSDC when compared with previously put clustering schemes Incremental Affinity Propagation clustering using K-Medoids and Incremental Affinity Clustering using Nearest Neighbour Assignment (IAPKM and IAPNA) give a comparable result for a streaming data clustering.