 |
|
|
|
|
|
| Intelligence Theories and Methods |
|
|
|
|
474 |
Risk Identification and Governance of Science and Technology Resource Security: Model Construction Based on Intelligence Thinking Hot! |
|
 |
Lai Jingming, Qian Guiming, Wu Qiuyi, Min Chao |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.002 |
|
|
Influenced by the evolving international landscape, incidents concerning the security of science and technology (S&T) resources in China have increased. In this study, S&T resource security was analyzed, its conceptual connotations and research dimensions were clarified, and risk typologies were identified. Building on this foundation, the core objective of “safeguarding national S&T security and stabilizing the S&T system” was operationalized into three pathways: situational awareness, agile response, and resilience building. Accordingly, a “perception-response-governance” model is proposed herein, characterized by government leadership with multi-stakeholder participation. The proposed model extends the S&T security research to include the granularity of S&T resources and integrates intelligence-driven thinking into the study of S&T resource security. It advocates constructing an artificial intelligence-enhanced S&T resource security monitoring platform, aiming to provide theoretical recommendations for safeguarding national security and facilitating scientific and technological advancements. |
|
|
2026 Vol. 45 (4): 474-486
[Abstract]
(
0
)
HTML
(1 KB)
PDF
(2180 KB)
(
14
) |
|
|
|
487 |
Patent Transaction Prediction for Technology Transfer: Perspective Based on Graph Neural Networks and Corporate Portraits Hot! |
|
|
Qian Minghui, Zhao Mengchun, Wang Chi, Wang Mingyu |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.003 |
|
|
The acceleration of the transformation of scientific and technological achievements is a crucial initiative for implementing the innovation-driven development strategy and advancing high-quality economic and social development. In this study, a heterogeneous graph neural network model based on enterprise and patent profiles, Company-Patent Profile Relational Graph Convolutional Networks (CP-RGCN), is proposed to predict potential patent-transaction partners. The model is validated using patent data from the artificial intelligence field. Experimental results demonstrate that CP-RGCN outperforms baseline models (Support Vector Machine, Random Forest, Graph Convolutional Network, Relational Graph Convolutional Network, Graph Sample and Aggregate, and Graph Attention Network) in key metrics such as Mean Reciprocal Rank and Hits@K. A feature importance analysis reveals that integrating enterprise and patent profile features within a heterogeneous graph neural network framework significantly enhances prediction accuracy while improving model interpretability and reliability. The proposed CP-RGCN model improves the identification of potential technology transfer opportunities and the commercialization efficiency of scientific and technological achievements. Furthermore, this study provides valuable insights and references for the application of heterogeneous graph neural networks to broader domains and scenarios. |
|
|
2026 Vol. 45 (4): 487-503
[Abstract]
(
1
)
HTML
(1 KB)
PDF
(3320 KB)
(
5
) |
|
|
|
504 |
Classification Space and Dynamic Evolution of Technological Weak Signals from Cognitive Perspective Hot! |
|
|
Zhang Huiling, Xu Haiyun, Chen Liang, Wang Chao, Liu Chunjiang, Wang Haiyan |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.004 |
|
|
In this study, the cognitive characteristics and processes of weak signals in science and technology are systematically analyzed, a cognitive space for weak signals and a transformation framework model for multiple signal types oriented toward technology foresight are constructed, monitoring of evolutionary trends in weak signal themes is implemented, and an early identification of emerging frontier technologies is promoted. The connotations and characteristics of weak signals in science and technology are synthesized and analyzed. The cognitive characteristics of the identification of such weak signals are then decoded. By integrating cognitive and evolutionary perspectives, a signal cognitive space oriented toward technology foresight is constructed, enabling full-spectrum signal classification from a cognitive perspective. The characteristic differences and transformation paths among multiple weak-signal types are explored, and a“double-cone”model combining perception and cognition is developed for signal classification and transformation path analysis. A case study in the stem cell field validates the feasibility of the proposed signal classification space and evolutionary analysis model. Relying on the constructed classification space for weak technological signals, this model is used to analyze the transformation mechanisms and evolutionary paths between signals, capture weak signals in the innovation field, and utilize a classification system and multidimensional evolutionary analysis to achieve an early identification of weak signals and avoid signal loss, thereby enhancing the accuracy and recall rate of weak technological signal identification. It dynamically tracks state transitions and trends within the classification space, supporting the prediction of emerging frontier topics based on signal evolution patterns. |
|
|
2026 Vol. 45 (4): 504-519
[Abstract]
(
1
)
HTML
(1 KB)
PDF
(3881 KB)
(
6
) |
|
|
|
520 |
Weak Signal Perception Model of Technical Demand: Interdisciplinary Insights from Brachistochrone Curve Hot! |
|
|
Yu Hui, Wu Yunjing, Xia Wenlei |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.005 |
|
|
The perception of weak signals in technological demand functions as a critical sensing and advisory mechanism within the field of science and technology intelligence. It plays a pivotal role in understanding and forecasting emerging technological development trends, as well as informing timely technological interventions. This study proposes a weak signal perception model grounded in the physical principle of the brachistochrone and evaluates its effectiveness and robustness. First, the study examines the brachistochrone trajectory of technological development within a two-dimensional“quantity-innovation”technology space, thereby elucidating the characteristics of technological progress. Second, a multi-layered temporal technological space is constructed based on technological distance, revealing both the developmental states and spatial distribution patterns of technologies. Third, the brachistochrone curve is fitted within this space to establish a fastest-spindle field, designed to identify weak signals of technological demand. Finally, the proposed model is validated through comparative analysis with alternative curve-fitting approaches. Empirical results based on real-world data on technological development demonstrate the superior effectiveness and robustness of the proposed model relative to benchmark models, enabling more accurate and efficient detection of weak signals in technological demand. Moreover, owing to the inherently latent nature of weak signals, the long-term effectiveness of the proposed model requires further time-based validation. Overall, the proposed model demonstrates strong performance in perceiving weak signals and exhibits preliminary decision-making potential for identifying technological development trends within the domain of frontier science and technology intelligence. |
|
|
2026 Vol. 45 (4): 520-533
[Abstract]
(
2
)
HTML
(1 KB)
PDF
(2929 KB)
(
8
) |
|
|
|
552 |
Research-Front Identification Based on Multilayer Semantic Networks Hot! |
|
|
Liang Guoqiang, Qiu Xiaopeng, Huang Xu, Zhang Shuo, Zhang Zhihao, Lin Gege |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.007 |
|
|
To address the semantic deficiency issues present in conventional research-front identification methods, i.e., citation analysis, text mining, and machine learning, this study proposes research-front identification based on multilayer semantic networks. DeepSeek-V3 was employed to extract subject-predicate-object (SPO) triplets in abstracts to facilitate the construction of multilayer semantic networks. Subsequently, cross-layer neighbor entropy and burst strength indicators were utilized to detect the activity and emergence characteristics of research fronts, with reverse matching of key nodes performed to obtain semantic information. Finally, empirical analysis was conducted with reference to carbon-capture literature published between 2014 and 2024 to evaluate the effectiveness of this method. The results show that the proposed method successfully identified approximately 80 emerging nodes and their semantic relationships in the carbon-capture field. Since 2024, these semantic relationships have focused on topics including “soil carbon sequestration,” “climate change,” “straw return,” “forest carbon sequestration,” and “model,” thus indicating that these topics represent research frontiers in the carbon-capture field. The extraction of SPO triplets based on large language models improves the efficiency and quality of the entities extracted. The proposed method addresses the semantic deficiency issues inherent in conventional research-front identification approaches and identifies research fronts in a research field. |
|
|
2026 Vol. 45 (4): 552-565
[Abstract]
(
2
)
HTML
(1 KB)
PDF
(10913 KB)
(
6
) |
|
|
| Intelligence Technology and Application |
|
|
566 |
Method for Identifying Regional Technological Opportunities Based on Science-Technology Multilayer Network Hot! |
|
|
Bai Guangyong, Mao Jin, Bai Yun, Li Gang |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.008 |
|
|
Identifying technological opportunities has become a key driver of scientific and technological innovation, while regions serve as the key spatial arenas in which innovation resources are concentrated and technologies are deployed. To accurately discern regional technological opportunities and thereby foster regional innovation, we propose a region-centric method for identifying technological opportunities. The approach establishes science-technology linkages by mining patent-to-literature citations and constructs a science-technology multilayer network. Based on this network, we develop SAGE-TSN (SAGE for Tech-Sci Network), which is a GraphSAGE-based model that fuses regional and global features of technological nodes with their scientific and technological attributes. Subsequently, a link-prediction framework is employed to highlight regional technological opportunities. Experiments encompassing 10 Chinese cities show that, after rigorous training and benchmarking, the proposed method significantly outperforms multiple baseline models in terms of accuracy, recall, and other metrics. The identified opportunities exhibit high activity and strategic importance. The proposed method effectively enhances the accuracy and regional adaptability of technological-opportunity identification, thus demonstrating considerable potential for practical application and broader deployment. |
|
|
2026 Vol. 45 (4): 566-578
[Abstract]
(
1
)
HTML
(1 KB)
PDF
(2467 KB)
(
9
) |
|
|
|
579 |
Construction of Generalized “Technology-Function” Matrix Incorporating TRIZ Knowledge Hot! |
|
|
Xi Chongjun, Zhao Yajuan, Zhang Ting, Lyu Lucheng |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.009 |
|
|
As a core tool for patent analysis, the “technology-function” matrix is pivotal in revealing the mapping relationships between technical approaches and functional effects, identifying research and development hotspots, and predicting industrial gaps. This study proposes a method for constructing a generalized “technology-function” matrix that integrates patent and paper data while incorporating Theory of Inventive Problem Solving (TRIZ) knowledge to overcome the limitation of single-dimensional analysis in conventional tools. First, the “technology-function” matrix model is constructed based on the “problem-method” framework, thus expanding the functional dimension into a multilayered structure comprising research objects (object, category) and research objectives (issue-E1, problem-E2, question-E3) while classifying the technical dimension into a progressive system comprising theoretical level (theory), methodological level(method) , and technical level (technology). By introducing TRIZ theory, the highest level of the functional dimension adopts a classification system comprising parameter, structural, and resource attributes, while the highest level of the technical dimension corresponds to solution tools such as inventive principles, standard solutions, and effect libraries. In terms of construction methodology, large language models are utilized to extract technology-function pairs from patent specifications and paper abstracts. This realizes the hierarchical mapping of technical terms to inventive principles and functional terms to engineering parameters through semantic analysis, thus ultimately forming a quantifiable multilevel matrix. Finally, considering patent and paper data in three fields—micro electro mechanical systems (MEMS) seismic detectors, power battery thermal management and safety control, and intelligent driving perception and decision control systems—as examples, a case analysis of a generalized “technology-function” matrix is conducted, where 39 engineering parameters and 40 inventive principles of TRIZ theory are combined. The results show that this model effectively achieves cross-domain correlation between papers and patents as well as identifies explicit technology layouts while uncovering potential innovation pathways, thus providing decision support with both academic depth and industrial application value for the technological roadmap planning of strategic emerging industries. This study offers significant theoretical and practical implications for improving the national innovation system. |
|
|
2026 Vol. 45 (4): 579-595
[Abstract]
(
2
)
HTML
(1 KB)
PDF
(2221 KB)
(
10
) |
|
|
|
596 |
Identification of Innovation Opportunities by Coupling Dimensions and Laws Under the Topics Evaluation and Screening of Technological Prediction Network: Based on the Triple Complex Relationships Between Technological Knowledge Elements Hot! |
|
|
Wang Jinfeng, Wang Congxiang, Zhang Ke, Feng Lijie, Feng Yicheng, Wu Zhishuang |
|
|
DOI: 10.3772/j.issn.1000-0135.2026.04.010 |
|
|
Exploring the complex relationships between technological knowledge elements for technological innovation not only allows enterprises to accurately and efficiently predict the development trend of related technologies by integrating opportunities between different technological knowledge elements but also allows enterprises to quickly respond to market demand and plan in advance to achieve sustainable development. However, existing research on identifying opportunities for technological innovation lacks sufficient exploration and integration of the complex relationships between technological knowledge elements and relies heavily on upstream task results combined with domain knowledge. This results in a certain level of blindness and disorder in technological innovation, significantly reducing the efficiency and quality of technological research and invention. In response, this study proposes an identification path for dimension-law coupling innovation opportunities through topic evaluation and screening of technological prediction networks using triple complex relationship mining between technological knowledge elements. First, by mining the direct semantic co-occurrence relationship, coupled co-occurrence relationship, and citation relationship between technological knowledge elements and using a weight optimization algorithm, a technological prediction network integrating the aforementioned triple complex relationships was constructed. Second, a three-dimensional strategic coordinate model was used to evaluate and screen out the core immature popular topics in the technological prediction network. Subsequently, based on the extraction of core immature popular topics subnets, an element variation innovation theory was used to identify technological innovation opportunities from the perspective of dimension-law coupling. Finally, a case study was conducted using 3D printing technology as an example to demonstrate the feasibility of the method proposed in this study. This study aims to provide reference concepts for accurately predicting the correlations between technological knowledge elements and efficiently identifying opportunities for technological innovation. |
|
|
2026 Vol. 45 (4): 596-616
[Abstract]
(
2
)
HTML
(1 KB)
PDF
(8756 KB)
(
6
) |
|
|
|
