A novel patent technology characterization method based on heterogeneous network message passing algorithm and patent classification system

Patents are widely recognized as important data for generating innovation. Accurate and rational characterizing patent technology content is a prerequisite for applying innovation generation algorithms. Existing research widely employs classification codes that annotate the technology content of pat...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Expert systems with applications Jg. 256; S. 124895
Hauptverfasser: Chang, Zhi-Xing, Guo, Wei, Wang, Lei, Fu, Zhong-Lin, Ma, Jian, Zhang, Guan-Wei, Wang, Zi-Liang
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 05.12.2024
Schlagworte:
Clustering algorithm
Heterogeneous network
Message-passing algorithm
Patent characterization
Technology topic
Technology topic
Heterogeneous network
Patent characterization
Clustering algorithm
Message-passing algorithm
ISSN:0957-4174
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Patents are widely recognized as important data for generating innovation. Accurate and rational characterizing patent technology content is a prerequisite for applying innovation generation algorithms. Existing research widely employs classification codes that annotate the technology content of patents to represent patents. However, the oversight of technology similarity information within patent classification systems results in deficiencies in the accuracy and effectiveness of their representations. To fill this research gap, we analyze the hierarchical structure of patent classification systems to extract the technology similarity information embedded within them. Then, we propose a novel patent technology characterization method based on the heterogeneous network message passing algorithm, which integrates the technology similarity information in the classification code co-occurrence information and the patent classification system to obtain a more accurate patent characterization. Subsequently, several evaluation experiments were conducted to compare our method with typical existing methods. The results demonstrate that our method outperforms these methods in accuracy and effectiveness. Finally, we conducted a case study to validate the reliability and practicality of our approach. In summary, our method exhibited superior performance, thereby providing robust support for innovation generation methods based on patent characterization, with high application value and extension prospects.
ISSN:0957-4174
DOI:10.1016/j.eswa.2024.124895