基于机器学习的专利质量分析与分类预测研究——以区块链技术专利为例  

Research on Patent Quality Analysis and Classification Forecast Based on Machine Learning——Taking Blockchain as an Example

作　　者：符川川 陈国华[2] 袁勤俭[1] Fu Chuanchuan;Chen Guohua;Yuan Qinjian(School of Information Management,Nanjing University,Nanjing 210023,China;School of Engineering Management,Nanjing University,Nanjing 210029,China)

机构地区：[1]南京大学信息管理学院,江苏南京210023 [2]南京大学工程管理学院,江苏南京210029

出　　处：《现代情报》2021年第7期110-120,共11页Journal of Modern Information

基　　金：江苏省文化和旅游科研课题项目“数字技术驱动下江苏省文化和旅游深度融合的战略重点和政策框架研究”(项目编号:20ZD01)。

摘　　要：[目的/意义]本研究在一定程度上可以减少专利审查员的时间成本和主观性并提高业务素质,也可以为专利申请者的专利布局提供参考。[方法/过程]本研究在专利质量界定与指标体系构建的基础上,提出基于机器学习的组合模型用于专利质量的分析与分类预测,并以新兴产业的区块链技术专利为例展开研究。该模型由自组织映射、核主成分分析以及支持向量机3种方法组成,其实现过程包括两阶段。第一阶段通过自组织映射(SOM)对从国家知识产权局专利数据库收集的21496项区块链技术专利数据进行分析并界定专利质量类别;第二阶段,通过核主成分分析(KPCA)对专利数据进行降维降噪处理,经过处理的专利数据再由支持向量机(SVM)得出分类结果。[结果/结论]在对区块链技术专利质量分类模型进行训练后,基于3306项区块链专利历史数据来验证训练模型的性能,实验结果的匹配度达到87.26%。因此,本研究提出的组合模型能够有效地对专利质量进行分类与预测。[Purpose/Significance]This research,to a certain extent,could reduce the time cost and subjectivity of patent examiners and improve their professional quality.It would also provide reference for patent applicants'patent layout.[Method/Process]This research defined the meaning of patent quality and constructed a patent quality indicator system.It proposed a combination model based on machine learning for the analysis and classification prediction of patent quality,and took blockchain patents in emerging industries as an example to conduct the research.The model was composed of three methods:self-organizing map,kernel principal component analysis and support vector machine.Its realization process included two stages.At first stage,self-organizing map(SOM)was employed to analyze the collected 21496 blockchain patent data from the patent database of the State Intellectual Property Office and define the patent quality category;at second stage,dimension reduction and noise reduction processing on patent data were performed through kernel principal component analysis(KPCA),the processed patent data was then classified by the support vector machine(SVM).[Result/Conclusion]After training the blockchain patent quality classification model,historical data of 3306 blockchain patents was used to verify the performance of the training model and the matching degree of the experimental results reached 87.26%.Therefore,the combined model proposed in this research can effectively classify and predict patent quality.

关 键 词：专利质量 机器学习 区块链 专利质量分析 分类预测 

分 类 号：G255.53[文化科学—图书馆学]