In financial field, predicting the future price of an asset has always been a hot topic. There are mainly two existing methods: One is to model the trend of asset prices in price prediction. Therefore, this method inevitably has a lag at the inflection point of the asset sequence. The other is to mine market opinion information from the internet to predict the future direction of prices. The challenge with this approach is that unstructured data processing and analysis is difficult. Therefore, we propose a method for asset movement prediction based on SABR [3] model. On the one hand, the market's prediction of asset trends implied in options can be used to solve the hysteresis problem. On the other hand, options data is easy to process and analyze. In this article, we try to use a neural network model to capture the market's view of the future trend of assets hidden in the stochastic volatility surface generated by the stochastic volatility model and establish a mapping relationship with asset prices. The results show that our methods can effectively eliminate the lag of price prediction and improve the accuracy of the prediction.