Clustering of Streaming Data using STRAP Algorithm

Now a day’s data stream clustering is more active research area, used to discover useful information from continuously generated huge amounts of data. There are various clustering algorithms related to data stream have been developed and proposed to make clustering on data stream. Clustering is the method of arranging the objects in one set, such that objects in the single group are more related to each other than those in other clusters (groups). The clustering of data stream imposes various challenges that need to be solved; some of them are dealing with dynamic data, capable of performing processing on fast incoming objects, also capable to perform processing of incremental data objects, and ability to address time, memory and cost limitations. The proposed STRAP clustering algorithm extends the Affinity Propagation (AP) to handle evolving data stream. It combines the statistical change point detection test with Affinity Propagation. It ingredients a group of labeled data objects with group of exemplars for detecting a changes in the generative process underlying the dream data. Experimental results with state-of-the-art data stream clustering methods demonstrate the effectiveness and efficiency of the proposed method. The proposed semi-supervised STRAP algorithm increases the accuracy and decreases the percentage of outliers compare to existing system.