Analysis of the impact of social network financing based on deep learning and long short-term memory

The risk of peer to peer lending (P2P) platform is predicted based on text data on the Internet to avoid the risk of social network financing and improve the security of social network financing. First, the transaction and review text information of a third-party P2P platform are classified for the time series of emotional changes. Second, the Granger causal relation test is used to verify the correlation between the time series of emotional changes and trading volume. Finally, a long short-term memory (LSTM) forecasting model is proposed based on investors’ emotional changes to predict the trading volume of P2P platforms using emotional changes as a reference for social network financing to avoid risks. The results show that the value of Pearson correlation coefficient between the trading volume of P2P platforms and negative emotions is -0.2088, with a P value less than 1%, indicating a correlation between emotional changes and trading volume. The Pearson correlation coefficient between the predicted and actual values is 0.7995, whereas the mean square error is 0.2190 with a fitting degree of 0.6532. This shows that the LSTM forecasting model can accurately predict the trading volume of P2P platforms with good performance in comparison with other forecasting models.