| The new round of scientific and technological revolution and industrial change is accelerating,the pace of global innovation is accelerating,China is going through a critical period of transformation of development mode,economic structure optimization and transformation of old and new dynamics,and innovation is increasingly becoming the key to cracking development problems.As a carrier of technology and knowledge,the number of patent applications is one of the most commonly used indicators of innovation in innovation research,and China’s emergence as an innovation leader is also reflected in the number of patent applications.Over the past 40 years of reform and opening up,our intellectual property business has achieved world-renowned success,but it is also facing the problem of being "big but not strong".This is the focus of national policy support in recent years,and is also a benchmark for the future examination of patent applications.The identification and prediction of the value of massive patent data is gradually becoming a hot topic of research.By combing through the relevant literature and addressing the shortcomings of existing research in terms of research context,research object and research method,we use complex networks,machine learning and causal analysis as the theoretical basis to evaluate patent value and analyse the influencing factors,and then investigate the causal relationship between the cooperation pattern of patent applicants and patent value among the influencing factors.This paper selects 254,172 valid patent data in the field of construction from 2011 to 2020 from the China Intellectual Property Office as the research object,and selects the influencing factors of patent value from two aspects: the patent’s own technical characteristics and the applicant characteristics of the patent.The patent’s own technical features are selected from those that have been published and proven to be relevant to the value of the patent.The applicant features of the patent are selected comprehensively and scientifically by constructing a two-layer coupled applicant cooperation-patent citation network model,combined with the structural features of the network,for indicators affecting the value of the patent.On this basis,this paper firstly uses the Random Forest algorithm for patent value assessment,and conducts importance ranking and partial dependency analysis on the selected factors influencing patent value to further explore the correlation between each factor and patent value.Based on this we find that the number of patent applicants and the importance of the patent applicant are two very important factors affecting the value of a patent,and that the number of patents with the phenomenon of collaboration has been increasing year on year in recent years,indicating that with the advancement of technology and the flow of knowledge,people are increasingly inclined to teamwork.Therefore,this paper uses the propensity score matching method to explore the causal relationship between the cooperation pattern of patent applicants and patent value,to estimate the causal effect between the cooperation pattern of patent applicants and patent value,and to conduct matching effect tests,sensitivity tests and robustness tests on the experimental results to confirm the accuracy of the results of this paper.The three subfields with the highest patent output in the construction sector were also selected for heterogeneity analysis to confirm the generalisability of the results of this paper in the construction sector.Based on the experimental results,we have the following conclusions.Firstly,this paper uses the plain random oversampling method to convert the patent citation dataset into a balanced dataset,and uses the random forest algorithm to evaluate the patent value.This method has the advantages of being scientific,feasible,accurate and efficient than the traditional regression method and neural network method,and is suitable for the value evaluation and patent value prediction of a large number of patents.It was then found that the length of the patent disclosure was the most important factor influencing the number of citations,followed by the number of claims,the number of applicants and the importance of the applicants,with the remaining two factors,the number of patent citations and the number of words in the patent title,being less important and having less impact on the prediction results.Finally,a causal analysis of patent applicant cooperation patterns and patent value showed significant causal relationships between several types of cooperation patterns and patent value.We conducted robustness tests and sensitivity analyses using four other matching methods and bias-adjusted matching estimates to assess hidden biases due to unobservable factors,and the results obtained were significant and less variable,further demonstrating the accuracy of our experimental results.This paper conducts a patent value assessment and analyses the influencing factors to explore the causal relationship between patent applicant cooperation patterns and patent value.For patent inventors,it can provide guidance to patent inventors to identify the important influencing factors of a patent in the early stage of patent creation and tend to devote their efforts to the important factors,to select a more favourable mode of cooperation and to avoid a situation where the patent they have invested their efforts in creating is not granted later.For the state,it is of strong practical significance to conduct intelligent evaluation of patent applications during the patent application burst period,to facilitate patent examination,to identify high value patents and low value patents in advance,to further screen low value patents,to avoid investing excessive resources and efforts,and to avoid a series of negative problems brought about after grant. |
