科研管理

科研管理 ›› 2020, Vol. 41 ›› Issue (8): 220-228.

• 论文 • 上一篇    下一篇

基于知识组合理论的技术机会发现

张振刚1,2,3,罗泰晔1   

  1. 1.华南理工大学工商管理学院，广东 广州510640;
    2. 广州市大型企业创新体系建设研究中心，广东 广州510640;
    3.广东省科技革命与技术预见智库，广东 广州510640
  • 收稿日期:2019-01-17 修回日期:2019-08-06 出版日期:2020-08-20 发布日期:2020-08-19
  • 通讯作者: 罗泰晔
  • 基金资助:
    国家社科基金重大项目(18ZDA062，2018.11—2023.12)。

 Technology opportunity discovery based on knowledge combination theory

 Zhang Zhengang1,2,3, Luo Taiye1   

  1.  1. School of Business Administration, South China University of Technology, Guangzhou 510640,Guangdong, China; 
    2. Guangzhou Research Center for Innovation System of Large- Scale Enterprise, Guangzhou 510640, Guangdong, China; 
    3. Science and Technology Revolution and Technology Forecasting Think Tank of Guangdong Province, Guangzhou 510640, Guangdong, China
  • Received:2019-01-17 Revised:2019-08-06 Online:2020-08-20 Published:2020-08-19
  • Supported by:
     

摘要： 本文以2011—2015年间纳米能源领域的8184个专利为例，提出了一种基于知识组合理论的技术机会发现新方法。知识组合理论认为，创新源于知识元素的组合，具体包括对已有知识元素组合的再利用和探索新的知识元素组合两种方式。一方面，本文应用关联规则对专利数据进行挖掘，用信息熵对所提取的关联规则进行筛选，从而得出具有较大再利用价值的知识元素组合；另一方面，本文提出了一个评估知识元素间新组合的可能性的公式，进而得出具有较大组合潜力的新知识元素组合。基于两种方式所得到的知识元素组合，我们分析出了纳米能源领域三个重要的研发方向，并以2017年7月至2018年6月间的专利数据验证了所提出的方法的有效性。

 

关键词: 知识组合, 技术机会发现, 关联规则, 信息熵

Abstract: Technology opportunity discovery refers to the process that identifies opportunities with potential application value by developing and utilizing technologies. It plays an important role in R&D organizations′ technology management. This paper aims to propose a novel method to discover technology opportunities based on knowledge combination theory. Knowledge combination theory holds that innovation emerges by combining or recombining knowledge elements in the processes of investigation and experimentation. Specifically, innovation derives in two ways: either from reconfiguring existing combinations so that they can be put to new uses or from combining knowledge elements in a novel manner. According to March′s work, the former way is called exploitative learning, and the latter exploratory learning. The two types of learning have different mechanisms. Exploitative learning is characterized by a process of refinement, selection and implementation based on organizations′ existing knowledge combinations. By contrast, exploratory learning denotes a conscious attempt to seek new knowledge combinations. So far, little research has taken the perspective of exploitative learning and exploratory learning to discover technology opportunities within specific fields. From the perspective of exploitative learning, existing knowledge combinations need to be assessed and those with great recombination value should be identified. From the perspective of exploratory learning, new knowledge combinations which may occur in the future need to be predicted. In many related studies, patent data is frequently used to predict the development trend of certain technologies. Patent analysis enables researchers to understand technology development directions and trends in an effective way. Therefore, in this paper, we take 8184 patents in nano-energy field from 2011 to 2015 as an example and try to discover technology opportunities from the two perspectives mentioned above. 

We first apply Stochastic Actor-Oriented Models (SAOMs) to test the validity of applying knowledge combination theory to innovative activities in the filed of nano-energy. Designed to model transitions in networks between discrete time points, SAOMs regard network evolution as nodes creating, maintaining or terminating ties to other actors. Modeling these kinds of structural processes allows researchers to understand the interactions among nodes in the networks. The statistical significance of the “degree act+pop” effect suggests that the theory is applicable in the filed of nano-energy. 
Based on the idea that exploitative learning aims to assess the reuse value of existing combinations of knowledge elements, we apply association rule analysis to the data set to find the knowledge combinations with high frequency. Association rule analysis is a classical method of data mining, which aims to determine frequent patterns, correlations or associations from various kinds of data sets. Each association rule yielded can be viewed as a knowledge combination. Given the large number of patents, we set the minimum support level at 1% and set the level of minimum confidence at 40%. A total number of 13 rules are yielded. To further assess the reuse value of the knowledge combinations, we introduce the concept of information entropy. Entropy represents the degree of technological diversification. It can be used to evaluate a knowledge element′s potential to interact with other knowledge elements. We calculate the average entropy of all the knowledge elements included in each association rule, and use it as an indicator of the rule′s reuse value. That is, the entropy of a knowledge combination is the mean of the entropy of all its knowledge elements. The higher the entropy, the greater the combination′s reuse value. The average entropy of the obtained rules is 17.37. We choose the four rules (i.e. knowledge combinations) whose entropy is greater than 17.37 for further analysis. We find that there is great reuse value for the development of nano-carbon materials and the application of nano-materials in semiconductor devices or active components. 
According to the idea that the purpose of exploratory learning is to discover new knowledge combinations, we construct a knowledge network to calculate the degree centrality of each knowledge element and the distance between knowledge elements. Based on these variables, we propose a formula to compute the combinatorial potential value between the knowledge elements which have no combinatorial history. We choose the ten combinations of knowledge elements with the largest combinatorial potential value for further analysis and find that it may become an important R&D direction to use various nano-materials in semiconductor devices to enable them to convert radiation energy to electric energy efficiently. In addition, applying nano-structured liquids or fluids to objects′ surface to achieve specific effects is also worth exploring. 
Combining the findings of two types of learning, we propose three major R&D directions in the field of nano-energy: the first one is the development and application of nano-carbon materials, such as the application of nano-carbon materials′ radiation and electrical properties; the second one is to explore the application of nano-materials in semiconductor devices; the third one is to apply nano-structured liquids or fluids to the surface of objects to obtain specific effects. To test the validity of the proposed method, we collect patent data in the field of nano-energy from July 2017 to June 2018. A total number of 2534 patents are found. We apply text association mining to the abstracts of the 2534 patents to yield keyword combinations related to the three directions mentioned above. The high occurrence frequencies of the keyword combinations provide support to the effectiveness of our method. 
This study has both academic and practical contributions. From an academic perspective, to our best knowledge, this study is one of the first attempts to apply network dynamics to analyzing the evolution of knowledge networks and find the combinatorial trajectory of knowledge elements. The estimation results of SAOMs provide support to the knowledge combination theory. From a practical perspective, this study proposes novel methods to identify technology opportunities from the angle of knowledge combination, which enriches the methodology of technology opportunity discovery based on patent information. The proposed method could help R&D organizations identify technology opportunities by utilizing publicly accessible data, based on which better decisions on research directions could be made.

Key words:  knowledge combination, technology opportunity discovery, association rule, information entropy

中图分类号: