【作者】 徐晟；

【导师】 汪应洛； 赵惠芳；

【作者基本信息】 合肥工业大学 ， 企业管理及其信息化， 2008， 博士

【摘要】 专利是一个区域科技资产的核心和最富经济价值的部分,专利的拥有量既能反映出该区域科技成果的原始创新能力,又能折射出这些成果的市场应用潜能,它是衡量区域创新能力和综合实力的重要标志之一。如何在区域创新的框架内对专利申请活动进行研究一直受到国内外学者的广泛关注,但是对专利申请驱动因素的研究并没有引起足够的重视。因此,文章选择专利申请的驱动因素作为研究对象,在理论上和实践上都具有重要的意义。文章在国家大力推进专利战略,加强知识产权战略,强化区域创新的大背景下,通过定性和定量相结合的方法深入研究了专利申请驱动因素。在研究的过程中,一方面运用支持向量机(SVM)方法对专利申请驱动因素进行建模,通过P-SVM和微分进化算法的混合模型选取非线性的关键驱动因素,并在负二项分布建模的基础上对关键驱动因素进行评价;另一方面运用实证分析的方法提出了中国专利申请驱动因素的相关假设,通过关系研究和验证性因素分析,研究了这些假设的存在性。文章主要研究内容和创新性成果如下:(1)文章通过对专利申请驱动的研究开发、经济增长、企业战略用途和专利保护程度等因素进行国际比较研究,分析了中国专利申请的驱动因素。文章以区域技术创新理论为核心,以专利申请活动为纽带,分析了中国区域专利技术创新体系的特征,提出了基于区域技术创新的中国专利申请的驱动体系,在专利申请驱动因素选取原则的基础上,分析了专利发展基础、专利发展环境、企业投入、政府投入、产学研联系和外商投资等中国专利申请的驱动因素及其变量。(2)文章在机器学习和统计学习理论基础上,引入支持向量机回归模型对专利申请进行建模。在支持向量机回归建模过程中纳入了区域经费及人力投入、专利倾向、区域富裕程度和外商投资活动等驱动因素变量,并通过仿真方法对专利申请进行预测。预测的结果一方面显示支持向量机预测方法比人工神经网络和逻辑回归方法有更高的预测精度,另一方面还表明运用支持向量机的方法对中国专利申请驱动因素的建模是可行和有效的,从而文章为处理小样本非线性的回归预测问题提出了一个可行的解决办法。(3)在通过支持向量机对专利申请建模成功的基础上,文章通过P-SVM和微分进化算法的混合建模对专利申请的关键驱动因素进行选取。在P-SVM建模的过程中对P-SVM的三个参数并行的进行了微分进化算法的优选,成功地选取了专利申请的关键驱动因素变量,如:区域GDP总量,每万人中大专以上学历人数,前期专利拥有总数,二产与三产从业人数之比,R&D人员数,企业R&D经费占销售收入之比,区域R&D总经费支出占GDP比重,大学和科研院所获得企业研究开发经费和人均外商直接投资额等。文章利用选取的关键驱动因素变量进一步对中国区域的专利申请进行预测,数据仿真显示具有良好特征抽取功能的P-SVM方法比核主成分分析的最小二乘支持向量机和岭回归方法具有较高的预测精度和推广能力。试验的研究表明P-SVM建模方法对处理类似专利申请驱动因素非线性回归问题的特征选取是有效的。(4)基于专利申请的泊松分布评价模型,文章结合中国专利申请数据样本的特征,提出专利申请的负二项分布的评价模型,并运用面板数据对中国东中西部地区专利申请驱动因素进行了评价、分析和讨论。分析结果表明:“区域GDP、前期专利拥有总数、企业R&D经费占销售收入之比、R&D总经费支出占GDP比重”这四个因素变量对东中西三个地区的专利申请都具有较强的驱动作用;“前期专利拥有总数、企业R&D经费占销售收入之比、R&D总经费支出占GDP比重”这三个因素变量对东中西三个地区的发明专利申请也具有较强的驱动作用,其他因素在不同地区针对不同的创新程度表现各异。(5)作为对专利申请驱动因素定量研究的补充,文章对中国专利申请驱动因素进行了实证研究。文章提出了专利申请与驱动因素之间的相关假设,利用各区域的调研数据,对专利申请与驱动因素之间的关系进行验证性因素分析,分析结果表示东部地区支持假设体系,中西部地区部分支持假设体系。文章针对中西部地区实证结论进行讨论,从专利发展的意识、法律环境、专利技术市场、自主知识产权及专利生命周期过程等方面提出了中西部地区专利发展、区域创新能力提高的对策建议。文章研究成果在一定程度上解决了专利申请驱动因素的非线性建模、特征抽取及评价的问题,为处理类似专利申请驱动因素非线性回归问题提供了更加广阔的空间。更多还原

【Abstract】 Patent is the most core and valuable part of a regional science and technology assets. The possession quantity of patents, which is the most important mark to describe the regional innovation ability and comprehensive strength, not only reflects the original innovation ability of this region, but also reflects the application ability of these achievements. It has been widely concerned that how to do the research in the framework of regional innovation. But the research of driven factors of patents application has not been paid enough attention. So it is very important both for theory and practice that the drivers are chosen as the study object.Under the background that national patents strategy vigorously promotes, intellectual property strategy improved, and the regional innovation strengthened, patent applications drive system is deeply studied using a combination of qualitative and quantitative methods. During this study, one side through support vector machine (SVM) model for the patent applications drivers, P-SVM is used to select the non-linear key driver, and negative binomial distribution model is used to evaluate the key drivers. Another side, the assumption of drivers of China’s patent applications is presented and the existence is analyzed by the relationship research and confirmatory factor analysis. The main content and innovative research results are as follows:(1) This article refers to the patent applications related to the driver of the theory, from the international comparative study of international research development, economic growth, corporate strategy and the degree of patent protection, a research perspective and the starting point are presented in this article, and the patent applications drivers are analyzed. In this article, the regional technological innovation is the core and the activities of patent applications are the link, and it is analyzed of China’s regional innovation system’s patented technology features. Driven system of patent applications is presented based on regional technological innovation. On the basis of the patent applications in the driver’s selection principles, the development of patents, patent development, business investment and government investment, research and contact with foreign investment drive, and other factors are analyzed.(2) In this work, on the basis of machine learning and statistics, support vector machine regression model is introduced for patent applications. Regional investment funds, manpower, patents tend to regional prosperity and foreign investment-driven activities are included in the modeling process as variable factors. The simulation methods of the patent application drivers to predict results show that support vector machines than the forecast of artificial neural networks and logistic regression methods have higher forecast accuracy, but also showed that the use of support vector machine approach to our patent application of the drivers Modeling is a feasible and effective in order to also deal with the non-linear regression prediction of a possible solution.(3) On the success of support vector machine modelling of the patent application, P-SVM is used in patent applications for selected key drivers. In the process of running P-SVM, for the P-SVM parameters, three parallel differential evolution algorithm for the selection of successful patent applications to select a key driver of variables, such as: the total regional GDP, per million people in college The number of qualifications, the total number of pre-owned patents, second and third production ratio of the number of practitioners, R&D staff, corporate R&D funds accounted for sales ratio, the total regional R&D expenditure proportion of GDP accounted for, universities and research institutes access to corporate R&D funding And the per capita foreign direct investment, and so on. Papers selected after the use of variable factors on China’s further regional patent applications to predict, data simulation shows a good feature extraction function of the P-SVM than kernel PCA least squares and support vector machine regression ridge of high forecast Precision and the ability to promote. The pilot study showed that P-SVM modelling approach in dealing with this kind of patent applications drive drivers return to the issue of non-linear feature selection to be effective.(4) Based on the patent applications in the evaluation of the Poisson distribution model, and considering our patent application of the characteristics of the sample data, a patent application for evaluation of the negative binomial distribution model adopted by the negative binomial distribution model for the evaluation of the use of panel data Eastern, central and western areas of patent applications drivers were evaluated, analyzed and discussed. The results show that: the regional GDP, the total number of pre-patent owners, corporate R&D funds accounted for sales ratio, R&D expenditures of the total proportion of GDP accounted for, the four factors of east, middle and west of China variable in the three regions of the patent application has a strong drive Role; the total number of pre-owned patents, corporate R&D funds accounted for sales ratio, R&D expenditure accounted for the total proportion of GDP. These three factors also east, middle and west of China variable in the three regions of the invention patent applications has a strong drive, other Factor in the different regions for different degree of innovation and performance varied.(5) The patent application for the quantitative study of the drivers added that empirical method is used for China’s patent application drivers. In this paper, the assumptions related to the patent applications and drivers are referred using of regional research data, patent applications and drivers of the relationship between the confirmatory factor analysis and the results indicated that the eastern region to support the assumption that the system, the central and western regions. To support the assumption that part of the system. The article for the central and western regions of the empirical conclusions were discussed, and the development of the patent awareness, legal environment, market, technology, independent intellectual property rights and patent life-cycle process, and other aspects of the central and western regions of the patent development, improve the regional innovation strategies suggestion.The research work in the dissertation have partly solved the problems of non-linear model for the patent applications drivers, feature extraction and evaluation, provided a broader space for the non-linear regression prediction, similer to the driven factors of patents application.更多还原