A Modified Approach For Incremental K-Means Clustering Algorithm

  Nidhi S. Shah

Clustering is process of grouping the object based on their attributes and features such that the data objects that are similar or closer to each other are put in the same cluster and it is form of unsupervised learning so no class labels are provided. K-means is most popular clustering algorithm which partitioned the data but there are many limitations of this algorithm such as number of clusters needs to be defined beforehand, number of iterations are unknown etc. Incremental data can be handled very efficiently by incremental clustering algorithm. It tries to generate new clusters at the end for each updated data, which cannot be merged with existing cluster so it increases computation time and also accuracy of clusters is reduced. This report presents most efficient modified approach for incremental K-means Clustering algorithm where clusters are generated dynamically without rerun the K-means on whole dataset which reduces computation time and gives more accurate result. For that, Initial Clustering is performed on static Database by using K-means clustering. Then for upcoming points, major distance between centroid to farthest point and upcoming point which is used to define the upcoming point is in existing cluster or not based on some criteria which is defined in proposed scheme. If it’s not in exiting cluster then recompute K-means for the outside points only.

Keywords- Modified Approach For Incremental K-Means Clustering

Unique Identification Number - IJEDR1502187
Page Number(s) - 1081-1084
Pubished in - Volume 3 | Issue 2 | May 2015
DOI (Digital Object Identifier) -   
Publisher - IJEDR (ISSN - 2321-9939)

  Nidhi S. Shah,   "A Modified Approach For Incremental K-Means Clustering Algorithm", International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.3, Issue 2, pp.1081-1084, May 2015, Available at :http://www.ijedr.org/papers/IJEDR1502187.pdf