Knowledge graph of mobile payment platforms based on deep learning: Risk analysis and policy implications

The Fintech mobile payment platform is expanding rapidly; this expansion, in turn, creates numerous risks. There is an urgent need to better understand these risks and to spur more secure payment behavior. This research aims to develop knowledge graphs of the mobile payment platform based on deep learning for risk analysis and policy inferences. We identify entities from collected policy documents, extract the relationships among the entities, and draw a risk knowledge graph on mobile payments. The use of unsupervised semi-automatic knowledge acquisition, we argue, can reduce the risk of mobile payment caused by a lack of knowledge. A significant benefit of this method is that risk knowledge can be acquired without supervision. Unlike other models, the absence of manual labeling allows for the relation extraction of triples to be unsupervised, while the previous triplet extraction was supervised. Compared with other unsupervised models, the precision of our model is improved, and the recall is the same as that of previous unsupervised shutdown extraction.Unsupervised relationship extraction can extract text relationships quickly and on a large scale, saving human resources for labeling.This method offers a potential solution to a fundamental problem; the content and quantity of policy documents exceed organizations’ and individuals’ ability to understand them. Our approach suggests the viability of developing a national policy risk knowledge graph to help mobile payment platforms understand national policies and reduce platforms’ operational risks while allowing users to quickly learn the risks of mobile payments and minimize the impact of those risks.