Patent analysis and classification prediction of biomedicine industry: SOM-KPCA-SVM model

Abstract

This paper proposed the application of a combinatorial model of machine learning to patent quality classification and forecasting in the biomedical industry. The model consists of three methods: Self-Organizing Map (SOM), Kernel Principal Component Analysis (KPCA) and Support Vector Machine (SVM), and names it SOM-KPCA-SVM model. The model proposed in this paper is implemented in two steps. First, the SOM groups the patent data and defines the patent level. Second, the patent data is reduced by KPCA to decrease noise, and SVM is applied to KPCA’s patent data to derive the classification results. The study collected 11,251 biopharmaceutical patent data from the patent transaction news. After training the patent quality model, 2196 historical patents were used to verify the performance of the training model. The accuracy of the match between experimental results and actual transaction status reached 84.13%. Therefore, the proposed patent quality method as a preliminary screening solution automatically and effectively evaluates the quality of patents. This method saves valuable time for reviewing experts, facilitates the rapid identification of high-quality patents, and can be used for the development of commercialization and mass customization of products.