A novel on automatic K value for efficiency improvement of K-means clustering

Abstract

The development of H/W and S/W has shortened the repetition cycle of new data generation and produced various categories of data. Machine learning, in particular, attracts explosive interest as it categorizes and analyzes data through artificial intelligence and contests against man. Once generated, data have their importance highlighted in terms of utilization. It is critical to analyze the data from the past and cluster new data for the utilization of data. The present study thus investigated an algorithm of determining the initial number of clusters automatically, which is part of problems with the K-means algorithm used in data clustering. The study also proposed an approach of optimizing the number of clusters through principal component analysis, a pre-processing process, with the input data for clustering. Its performance evaluation results show the accuracy rate of 87.6% or so.