A Novel Technique to Enhance K-Mean Clustering using Back Propagation Algorithm
Indian Journal of Science and Technology
Objectives: In this work, improvement in the k-mean clustering is proposed in terms of accuracy and execution time. Methods/Statistical Analysis: The clustering is the technique which is used to analyze the data in the efficient manner. In the recent times various clustering algorithms has been proposed which are based on different type of clusters. Among the proposed algorithms k-mean performs batter in terms of accuracy and execution time. In the k-mean clustering algorithm, the dataset is
... , the dataset is loaded and from which number of attributes and members are analyzed. The arithmetic mean is calculated from the dataset which will be the central point. The central point is considered as the referral point and Euclidian distance to calculate to all other points in the dataset. The Euclidian distance is calculated by taking single central point due which accuracy of clustering is reduced. In the work, back propagation technique is applied which will calculate Euclidian distance multiple times and achieve maximum accuracy of clustering. Findings: The k-mean clustering is the efficient clustering technique which clusters the similar and dissimilar type of data. The k-mean clustering calculates data similarity by calculating the Euclidian distance from the central point. The central point is calculated by taking arithmetic mean of the dataset. When the dataset is complex and large in size, k-mean clustering is not able to drive exact relation between the data points due which some points are left uncluttered. The back propagation algorithm is used to drive accurate relationship between the data points. In the back propagation algorithm input values and actual clustering is derived and this process continues unless maximum accuracy is achieved of clustering. The proposed algorithm is implemented in MATLAB and it is being analyzed that accuracy is increased upto 15 percent and execution time is reduce 2 seconds as compared to existing k-mean algorithm.