Abstract: From the perspective of public opinion analysis, an online lending problem platform identification model is built based on experimental research, combining internal platform information and external public opinion information. The model is constructed by using neural network, support vector machine, random forest and logistic regression in order to verify the role of public opinion information indicators in improving the performance of the identification model. The results show that, firstly, the accuracy of comprehensively using fundamental information, operation information and public opinion information to identify the online lending problem platform is higher; secondly, compared with the random forest model, the support vector machine model and the logistic regression model, the identification effect of the neural network model is the best. Therefore, the study is helpful for regulators to understand the platforms operating status and to carry out targeted regulatory governance for different platforms scientifically and comprehensively.

Key words: online lending, online public opinion, subject extraction, machine learning