Study on Classification of Patents Collaborative Filtering Oriented to TRIZ
Hu Xuegang1, Yang Hengyu1,2, Lin Yaojin3, Bao Yanwei1
1. School of Computer and Information, Hefei University of Technology, Hefei 230009; 2. Anhui Institute of Scientific and Technical Information, Hefei 230011; 3. School of Computer Science, Minnan Normal University, Zhangzhou 363000
Abstract With increasing applications and attention to patents, the study of patent classification has attracted wide attention. As an important technology of recommendation systems, collaborative filtering technology is widely used in the field of Internet data mining, and has the characteristics of simplicity and efficiency. In this paper, a patent TRIZ classification method based on collaborative filtering (TRIZ-CF) is proposed based on the idea that patent inventors use similar methods to solve similar problems. First, the method quantifies the patent text. Second, a patent scoring matrix is constructed. Finally, unclassified patents are classified by the calculated ranking scores. Experimental results show that the proposed method can improve classification accuracy compared to SVM (Support Vector Machine) and efficiency and can be effectively applied to TRIZ-CF.