Discovering Technological Opportunities in the Hydrogen Industry: An Adoption of GNN and Patent Data

It is increasingly important to timely discover technology opportunities in nowadays business, while patent data has been widely used for technology opportunity discovery (TOD) so far. However, most available researches focus on intricately processed one-dimensional co-occurrences or citations, as the information hidden in higher dimensions, such as patentee-technology relation, has been ignored so far. To fetch this valuable information, this paper adopts a graph neural network method with a recursive process, for analyzing the internal capacity of firms, and combines it with external challenges evaluated by competitiveness index. Taking the hydrogen industry with 39,413 patent families from 1969 to 2021 as the basic dataset, the case study shows a 32.49% promotion compared to the analogous TOD model. The proposed opportunities are tested by the recent filings of the case entity, Dalian Institute of Chemical Physics. This approach provides a novel perspective for patent data mining and contributes theoretically to the network analysis method.