An efficient grid algorithm for faster clustering using K medoids approach

Abstract

Clustering is the methodology to separate similar objects of data set in one cluster and dissimilar objects of data set in another cluster. K means and K medoids are most widely used Clustering algorithms for selecting group of objects for data sets. k means clustering has less time complexity than k medoids method, but k means clustering method suffers from extreme values. So, we have focused our view to k medoids clustering method. Conventional k-medoids clustering algorithm suffers from many limitations. We have done analysis on these limitations such as the problem of finding natural clusters, the dependency of output on the order of input data. In this paper we have proposed a new algorithm named Grid Multidimensional K medoids which is designed to overcome the above limitations and provide a faster clustering than K medoids..

Performance Evaluation of Selected Distance-Based and Distribution-Based Clustering Algorithms

“…The algorithm was found to be very efficient and showed better tolerance than the k-means. The k-medoids is widely proposed in several other studies (Daiyan et al, 2012;Joshi et al, 2011;Reynolds et al, 2004;Zadegan et al, 2013). In order to optimize the performance of k-means, the algorithm was modified in (Daiyan et al, 2012), the modification resulted in a faster clustering and better cluster quality.…”

“…The k-medoids is widely proposed in several other studies (Daiyan et al, 2012;Joshi et al, 2011;Reynolds et al, 2004;Zadegan et al, 2013). In order to optimize the performance of k-means, the algorithm was modified in (Daiyan et al, 2012), the modification resulted in a faster clustering and better cluster quality. In a related study proposed in (Ding et al, 2007), the authors in the course of their research reported the use of k-means for the generation of class labels.…”

“…The object of the same cluster shares some features, and by the perspective of machine learning, clusters directly correspond to hidden patterns (Berkhin, 2006). K-means algorithm groups data based on distance measures; it is the most reported clustering algorithms for partitioning of data sets into group of objects (Daiyan et al, 2012). Also, survey studies reported in (Berkhin, 2006;Fahad et al, 2014) showed that, this technique is faster and describe it as the leading distance-based clustering technique.…”

Cluster Analysis of Data Points using Partitioning and Probabilistic Model-based Algorithms

“…The algorithm is widely proposed in several other studies [12], [5], [13] and [11]. The algorithm was modified in [5] to get a faster clustering and to overcome some of its limitations…”

“…The clustering algorithm automatically supplies clusters found in data with a conceptual description and according to [4], a good conceptual description can be used for better understanding and better decision. K-means and k-medoids are the most widely used clustering algorithms for selecting group of objects from data sets [5]. Traditionally, clustering techniques are broadly divided into hierarchical and partitioning [3], the notion used in both techniques to cluster data defers.…”