LAUSR

In order to help financial users to invest, and to provide users with comprehensive and accurate information about financial securities, information about financial securities from multi-heterogeneous information is obtained. The characteristics of financial information are analyzed to provide valuable investment advice to users. According to the financial characteristics of the user’s interest, the characteristics of the investor’s interest are extracted from the heterogeneous information. Then, the multi-level model is proposed to analyze the characteristics. Through the multi-level model, the conversion of convertible bonds and the net value of closed funds are predicted. In the first level, based on the characteristics of convertible bonds and closed funds, three models of trend evaluation model, SVR (Support Vector Regression) model and neural network backpropagation network (BPN) model are used to predict financial characteristics. In the second level, the results produced by the three models in the first level are fused by the neural network. The third level optimizes the neural network based on the second level. The optimal initial weights and thresholds are selected by genetic algorithm to obtain better prediction results. The results show that the model can predict the characteristics of convertible bonds and closed funds more accurately. Therefore, the model provides a certain reference for financial users’ investment.