I 摘要 在如今知识经济高速发展的时代背景下，专利在将科技优势转化为竞争优势 的过程中意义重大，在促进人类社会科技进步及经济发展中作用显著。近些年国内 外频发的专利热点事件，无不在对企业传达着信号：掌握核心技术、获取专利权才 是市场竞争中的硬实力。因此，各国开始在从技术研发到获取专利的各环节上制定 战略，加大资源投入，致使专利数量猛增。而与此同时，专利增速过快带来的专利 泡沫也为专利正常发挥作用带来阻碍。为了促进专利正向发展，保障优质专利有效 发挥其应有的作用，对专利质量的有效评估意义重大。 现有对专利质量的研究总体上可以概括为两类：第一类是将专利质量作为影 响群体创新或企业绩效的前因，使用单一指标作为专利质量的代理变量；第二类是 将专利质量作为结果，从宏观或微观层面来对影响专利质量的若干因素进行分析， 并在此基础上探究专利的评估模型及方法。但目前对专利评估的研究大多局限于 对评估指标体系的理论构建，缺乏实证评估结果的支撑。随着对知识产权重视程度 的提升，近年来国内专利数据库建设取得重要进展，各类专利相关数据资源日趋丰 富，为专利评估的实证研究提供了良好条件。因此，本文在系统梳理、归纳、对比 现有专利评估研究的基础上，构建了一个更加综合性的专利质量评估指标体系，并 通过对比和分层抽样构建一个专利数据集，用两种方法对该指标体系的评估效果 进行实证研究，得到相应的结论，以期为我国专利质量评估研究和实践工作提供参 考和借鉴。 本文搜集整理中国专利奖4089件获奖专利数据，并通过分层抽样获取4089件 非获奖专利作为对照，组成专利质量评估训练集，同时参考我国高新技术的产业划 分及国民经济行业的划分，将训练集数据分成包括化学、电子信息等六类。运用实 证方法，首先使用T检验与卡方检验对指标进行初步筛选，进而对不同训练集使 用逻辑回归与判别分析两种方法来确定对应的专利质量评估指标，建立评估模型， 对评估模型的稳健性进行检验。在此基础上，利用建立的评估模型对石墨烯专利测 试集进行质量评估，分析比较评估效果，得到以下两个主要结论： 1）家族被引次数、发明人个数、实质审查时长、扩展同组个数、专利许可和 质押这六项指标是专利质量评估的核心评估指标，专利优先权个数、申请人类型、 专利诉讼情况为辅助指标； 2）专利质量评估模型的有效性具有领域差异。因为不同领域知识结构和创新 路径存在差异，本文认为不同领域专利质量评估模型指标的有效性存在差异，并得 摘要 到数据分析结果的支持。 本文提出的专利质量评估模型基于客观指标，具有数据可得性强、评估效果稳 定的特征，是对现有专利质量评估研究的补充和完善。论文进一步针对不同产业领 域，提出了差异化的评估体系，具有更强的针对性和操作性。 关键词：专利质量，评估体系，逻辑回归，判别分析，专利奖 ABSTRACT III ABSTRACT In the context of the rapid development of the knowledge economy today, patents are of great significance in the process of transforming technological advantages into competitive advantages, and play a significant role in promoting human society’s technological progress and economic development. In recent years, at home and abroad， frequent patent hotspot events have conveyed signals to enterprises that mastering core technology and obtaining patent rights are the hard power in market competition. Therefore, countries began to formulate strategies in each link from technology R&D to obtaining patents and increase capital investment, resulting in a sharp increase in the number of patents. At the same time, the patent bubble brought about by the rapid growth has also hindered the normal functioning of patents. In order to promote the positive development of patents and ensure that high-quality patents effectively play their role, the effective evaluation of patent quality is crucial. The existing research on patent quality can be summarized into two categories: the first category is to use patent quality as a cause of group innovation or corporate performance, using a single indicator as a proxy variable for patent quality; the second category is to classify patent quality as a result, several factors affecting the quality of patents are analyzed from the macro or micro level, and on this basis, the patent evaluation model and method are explored. However, the current research on patent evaluation is mostly limited to the theoretical construction of the evaluation index system, lacking the support of empirical evaluation results. With the increasing emphasis on intellectual property rights, important progress has been made in the construction of domestic patent databases in recent years, and various patent-related data resources have become increasingly abundant, providing good conditions for empirical research on patent evaluation. Therefore, on the basis of systematically combing, summarizing, and comparing existing patent evaluation research, I build a more comprehensive patent quality evaluation index system, construct a patent data set through comparison and stratified sampling, use two methods to conduct empirical research on the evaluation effect of the index system, and get the corresponding conclusions, with a view to providing reference for China's patent quality evaluation research and practical work. This article collects and organizes the data of 4,089 winning patents of the Chinese ABSTRACT Patent Award, obtains 4,089 non-winning patents as a control by stratified sampling to form a training set for patent quality evaluation, and divides data into six categories including chemical, electronic information, etc. By using empirical methods, firstly I use T test and chi-square test to initially screen the indicators, and then use two methods of logistic regression and discriminant analysis on different training sets to determine the corresponding patent quality evaluation indicators, establish an evaluation model, and test the stability of the evaluation model. On this basis, I use the established evaluation model to evaluate the quality of the graphene patent test set, analyze and compare the evaluation results, and obtain the following two main conclusions: 1)The six indicators of family citations, the number of inventors, the length of substantive examination, the number of extensions in the same group, patent licensing and pledge are the core evaluation indicators of patent quality evaluation; the number of patent priority, the type of applicant,the situation of litigation are auxiliary indicators; 2)The validity of the patent quality assessment model differs by domain. Because of the differences in knowledge structure and innovation paths in different fields, I propose a hypothesis of “the difference in the validity of the indicators of patent quality assessment models in different fields” and the hypothesis is supported by data analysis results. The patent quality evaluation model proposed in this paper is based on objective indicators and has the characteristics of strong data availability and stable evaluation effect. It is a supplement and improvement to the existing patent quality evaluation research. The thesis further proposes a differentiated evaluation system for different industrial fields, which is more targeted and operational. Keywords: Patent quality, Evaluation system, Logistic regression, Discriminant analysis, Patent award 目录 V 目录 第一章 绪论 .................................................................................................................... 1 1.1 研究背景 ........................................................................................................... 1 1.2 研究问题 ........................................................................................................... 3 1.3 研究意义 ........................................................................................................... 4 1.4 研究方法 ........................................................................................................... 5 1.5 研究框架 ........................................................................................................... 6 1.6 结构安排 ........................................................................................................... 7 第二章 文献综述与理论基础 ........................................................................................ 9 2.1 专利质量梳理 ................................................................................................... 9 2.1.1 专利质量的概念