Yapay Zeka Hızlı Bakış & Çıkarımlar
AI InsightAmong them, k-nearest neighbors demonstrated the highest performance.
📌 3 Maddede Bu Makale
Abstract
A patent contains various information about a developed technology and is a form of Big data that receives millions of applications worldwide each year. Recently, there has been an increase in research that analyzes such patent Big data for use in R&D strategy establishment. Among these studies, a core patent classification is recognized as important because it can be used for a variety of management information. In the past, the core patent classification was performed qualitatively by some experts, but it was expensive and time consuming. To complement qualitative methods, quantitative methods using statistics and machine learning are being studied. Existing proposed methods utilize the quantitative indicators specified in the patent. However, quantitative indicators have different values for each elementary technology. If this characteristic is not reflected, an incorrect analysis result is produced. In addition, various values such as rights, technology scalability sustainable development, etc., must be considered in order to effectively classify core patents. In this paper, we propose an effective core patent classification model using improved patent performance indicators. The proposed model applies text mining and clustering to patent Big data to identify elementary technology and calculate improved patent performance indicators that reflect various values. Furthermore, a core patent classification model is constructed by learning various classification algorithms. In order to examine the practical applicability of the proposed model, experiments are conducted with patents registered in the USPTO. As a result of the experiment, the accuracy of three models trained with patent-improved performance indicators was high. Among them, k-nearest neighbors demonstrated the highest performance.