From technology opportunities to solutions generation via patent analysis: Application of machine learning-based link prediction
Ziliang Wang,Wei Guo,Hongyu Shao,Lei Wang,Zhixing Chang,Yuanrong Zhang,Zhenghong Liu
DOI: https://doi.org/10.1016/j.aei.2024.102944
IF: 8.8
2024-11-22
Advanced Engineering Informatics
Abstract:Technology convergence represents a significant mode of technological innovation that is widely prevalent across various industries. This innovative approach integrates multiple technologies to develop new integrated solutions, thereby fostering a competitive advantage for enterprises. Anticipating future potential technology convergence is of paramount importance for businesses. However, previous research has predominantly relied on the topological information of convergence networks, overlooking the nodal attributes and inter-nodal relationships that have an impact on the emergence of technology convergence. To enhance existing studies, this paper employs three types of features: node attributes and inter-node relationships based on the drivers of technology convergence, along with link prediction similarity indices. Additionally, we utilize Graph Convolutional Neural Network (GCN) for node embedding to leverage node attributes. Machine learning models are utilized for link prediction based on these features to identify potential technology opportunities. To guide research and development (R&D) efforts, we recommend high-value patents for each node using entropy weighting across five metrics that objectively quantify patent value, and transform patent abstracts into vectors using Doc2Vec. Patents with high similarity in abstract text between nodes are utilized to extract technical solutions and fuse ideas for technology convergence. A case study is conducted within the autonomous driving industry, leveraging comprehensive information including node attributes, inter-node relationships, and topology-based similarities to identify technology opportunities and guide the generation of R&D ideas through the convergence of technical solutions.
engineering, multidisciplinary,computer science, artificial intelligence