The changes of the stock market affect the returns of investors, but the trend of stock prices is affected by many factors, and the traditional forecasting methods are increasingly difficult to meet people's requirements. Considering that CNN is good at processing images and LSTM is good at predicting events in time series, we propose a model based on ResNet and LSTM. Firstly, we use a network structure similar to 18-layer ResNet to extract features from the data. Then LSTM is used to predict stock prices. Experimental results show that this model can be successfully applied to the research of stock prediction, and it is more stable than other methods.