Scientific and technological innovation is a powerful driving force for high-quality industrial development. This study proposes a frontier technology identification method based on artificial intelligence agents (AI Agent) and multi-feature fusion of text, providing reference for grasping technological breakthrough directions and exploring technological innovation opportunities. Taking fund projects, papers, and patent data as research objects, first, AI Agent is used for data preprocessing to extract text keywords; secondly, semantic features, classification features, and Biterm topic model topic features of scientific and technological literature are fused to represent text vectors, enhancing data features, and clustering algorithms are used to divide technical topics; finally, a frontier technology identification index system based on novelty, growth, influence, and innovation characteristics is constructed to identify frontier technologies. An empirical study in the field of micro light-emitting diodes shows that the clustering method integrating the three types of features (semantic, classification, and topic) performs well, outperforming methods based on a single feature or two features. It identifies emerging frontier technologies such as chip manufacturing technology and full-color technology; hotspot frontier technologies such as luminescent materials and devices, display screen technology, holographic display, medical applications, detection and repair, and driving circuits; as well as potential frontier technologies such as massive transfer technology, proving the effectiveness and application value of the method.
关键词
前沿技术识别/AI Agent/多特征融合/Biterm主题模型
Key words
Frontier Technology Identification/AI Agent/Multi-feature Fusion/Biterm Topic Model
