A comparison of patent classifications with clustering analysis

Abstract

There is an abundance of data and knowledge within any given patent. Through the use of textual mining and machine learning clustering techniques it is possible to discover meaningful associations throughout a corpus of patents. This research demonstrates that such relationships between USPTO patents exist. Through the use of k-means and k-medians clustering, the accuracy of the USPTO classes will be assessed. It will also be demonstrated that a more refined classification process would be beneficial to other areas of analysis and forecasting.