A technology forecasting model using support vector clustering and voting approach

Abstract

Patent clustering refers to assigning patent documents to similar clusters using document clustering techniques. General document clustering is based on traditional clustering algorithms, such as K-means and hierarchical clustering. The number of clusters is determined for the general clustering algorithms. Patent document clustering also requires the optimal selection of the number of clusters. Much research has been introduced for determining the number of clusters. However, most of the research has focused on numeric data The optimal selection of the number of clusters is equally important to patent clustering, but it is difficult to determine the number of clusters in the patent document data because the documents have diverse data types, such as text, number, and picture. Thus, we need another approach to determining the number of clusters for patent document clustering. In this paper, we propose a method for determining the number of clusters in patent document clustering, using support vector clustering. This research will also develop a new method for evaluating the determined number of clusters using a voting approach of an ensemble method. To verify the performance of this research, we will carry out an experiment using patent document data from Korea Intellectual Property Rights Information Service and The U.S. Patent and Trademark Office. ©2013 International Information Institute.