Clustering is a way of combining data objects or data points into disjoint cluster. The basic concept behind clustering is that the data objects in the same clusters should be related to each other and the data objects belonging to different clusters should differ from each other. This research paper proposes a new algorithm which combines the features of K-means clustering algorithm and Hierarchical clustering algorithm BIRCH. The proposed algorithm first perform hierarchical clustering on the dataset which gives a large number of clusters and then further perform partitioning clustering using K-Means partitioning clustering algorithm to reduce the number of clusters and get more accuracy. The proposed algorithm is applied on cars dataset which is then compared with K-means clustering algorithm. The comparison is done on the basis of within sum square error in which the new algorithm give better results as compare to K-Means clustering algorithms.

No abstract

Abstract

Data mining is a method of finding useful patters from large volumes of data. It is an extension of traditional data analysis and statistical approaches. Data Clustering is a task of grouping a set of items or objects into subsets (called clusters). It is an algorithm to discover the similarity between objects in the same class (intraclass similarity) and minimizing the similarity between objects of different classes (interclass similarity). This paper discusses the standard KMeans clustering algorithm and Kohonen Self Organizing Map(SOM) clustering algorithm using the Tanagra datamining tool .These algorithms are applied on facebook dataset i.e which type of information is shared by university students on facebook.And that information is then used for product marketing purposes. And according to our analysis SOM gives best result with high accuracy and less computational time.

No abstract