The impact of the investor sentiment on China’s capital market price volatility is concerned under the perspective of the behavioral finance. Firstly, in terms of the existing methods of establishing the investor sentiment index, the composite investor sentiment index which include six indicators (five objective indicators and a subjective indicator) are obtained. Secondly, VMD-LSTM (Variational Mode Decomposition and Long Short Term Memory) hybrid neural network model is used to decompose and restructure the investor sentiment index and the Shanghai Security Exchange Composite Index (SSEC) into the short-term, medium-term and long-term trend. Each trend is trained to obtain the forecasting results in three different time scales, and then to achieve the final predicting results by superimposing the output of each trend. Furthermore, compare with other prediction methods, the model can indeed improve the overall predicting accuracy. Finally, GARCH model and the co-integration error regression model are used to discuss the fluctuation correlation and VAR (Vector Auto-regression) models are established to analyze the causality between the stock market indices and the investor sentiment index.