Digital Currency Risk Knowledge Acquisition Model Based On Machine Learning

The risk of digital currency in the big data scenario lies not only in the digital currency itself,but also in the complex and heterogeneous trading platform and the difficulty in acquiring risk knowledge for investors.Based on the perspective of knowledge management,this paper obtains digital currency risk knowledge from the trading platform and investors.In order to obtain the behavior features of investors in digital currency transactions,the comments data of investors in digital currency forums were retrieved,and BERT,a natural language processing model proposed by Google,was used to make sentiment classification of the comments.Based on the attention mechanism,the feature words with high weight were obtained,and AUC was used as the evaluation index of sentiment classification.Secondly,the trained BERT + LDA is used to extract the subject and corresponding keywords of the comments,and analyze the behavior features of investors in digital currency trading.For the risk features of the digital currency trading platform,crawl trading platform features and risk event data from non-small,bitcointalk.org and Bitcoin Wiki websites,using XGBoost to learn to identify high-risk trading platform of data,it is concluded that the risk features of digital currency trading platforms importance,AUC as evaluation index,and verify the reliability of digital currency trading platform important risk features.Based on the framework of knowledge management,namely the paradigm of data-information-knowledge-wisdom,the knowledge is summarized and extracted from the features of investors’ behavior and trading platform,and the acquired knowledge is classified from three dimensions of people,organization and technology.The contribution of the paper is as follows: the paper is based on the epistemology and the theory of behavioral finance research digital currency risk,on the dimension of the data source,and the risk of get knowledge from basic information theory knowledge management perspective,to get more granular digital currency trading risk knowledge,and fill the gaps in the theoretical research on digital currency risk,and digital currency trading has a guiding significance for investors.The investor risk knowledge acquisition model based on BERT+LDA designed in this paper acquired new knowledge,and the maximum KL based on BERT + LDA model was 8.98,which was greater than the maximum KL of Lda2 vec of 8.32;Compared with the Lda2 vec model,the model proposed in this paper identifies new topics of technical discussion,which shows that the attention to digital currency technology has an impact on the risks of digital currency transactions.Key words under the topic of technical discussion can provide think tank suggestions for investors.Trading platform based on XGBoost risk identification model can better identify high-risk trading platform,the AUC value is 0.83,higher than that of other classification algorithms,trading platform by the number of hackers,activity and daily trading volume of investor choice trading platform should focus on factors.