To search, Click below search items.


All Published Papers Search Service


Cluster Feature-Based Incremental Clustering Approach (CFICA) For Numerical Data


A.M.Sowjanya, M.Shashi


Vol. 10  No. 9  pp. 73-79


Data clustering is a highly valuable field of computational statistics and data mining. A major difficulty in the design of modern data clustering algorithms is that, in majority of applications, new data sets are dynamically appended into an existing massive database and it is not viable to perform data clustering from scrape every time new data instances get added up in the database. The development of clustering algorithms which handle the incrementally updated data points, has received increasing interest among the researchers. This paper presents a more efficient cluster feature-based incremental clustering approach (CFICA) for numerical data sets. Initial clustering is performed on the static database with the help of k-means clustering algorithm. Then, by making use of cluster feature computed from the initial clusters, the incrementally updated data points are clustered. Subsequently, the closest pair of clusters is merged to obtain better cluster accuracy. Finally, the proposed approach has been validated with the help of real datasets presented in the UCI machine learning repository. The experimental results demonstrated that clustering accuracy of the proposed incremental clustering approach is improved significantly.


Data mining, Clustering, Incremental clustering, k-means algorithm, Cluster Feature, Mean, Farthest neighbor points, Clustering accuracy (CA)