Research On Risk Identification Of Equity Crowdfunding Based On Deep Learning

Internet equity crowdfunding,as a mode of Internet financial crowdfunding,has provided new channels for the financing of small and medium-sized micro enterprises,improved the cooperation efficiency of both sides of investment and financing,changed the current financing environment of small and micro enterprises in China,and provided impetus for the development of entrepreneurial economy.Because Internet equity crowdfunding is in the initial stage of integration,there are still a lot of risk problems.Based on this situation,this paper takes the Internet equity crowdfunding project as the research object,from the perspective of investors,hoping to provide new basis for the risk control of all parties,especially investors,by studying the risks faced by relevant participants.Before writing this paper,on the basis of literature research on Internet equity crowdfunding in recent years,through the analysis of the development of Internet equity crowdfunding in recent years,we found that all parties have risk problems in varying degrees.By using the risk tree search method,we can find out the project initiator,the project itself,the project protection mechanism and the later stage of the project Based on the in-depth analysis of the four aspects of exhibition and the actual situation,21 risk identification indicators of Internet equity crowdfunding are preliminarily established.Secondly,due to the dynamic volatility of Internet finance,in combination with the experimental data captured by Python code,and in the case of high integrity of the original data and sufficient data for the experiment,after a series of data processing,such as eliminating invalid data,data index integration processing,continuous data standardization processing and so on,we finally get the risk control types and provinces that can be provided 17 risk identification indicators,such as the types of information available,are used to build the risk identification indicator system of Internet equity crowdfunding.Then,we use the pytorch framework to build the deep neural network model.After training and testing the model has reached a certain accuracy,we can identify the data that is not labeled by the model and draw a preliminary research conclusion.The result shows that the recognition accuracy of the model reaches 98%,which confirms the applicability of the deep neural network in the Internet equity crowdfunding risk research.The cross validation with 10% discount is used to prove the scientificity,objectivity and preciseness of the empirical study.Finally,the importance index is obtained by catboost algorithm,and the key factors affecting the risk of Internet equity crowdfunding project are obtained.