In China, stock investment is o ne of the important ways for people to manage their finances. Mastering the stock market dynamics can not only bring economic benefits to individuals and enterprises, but also help government to understand Chinese macroeconomic situation. The overall situation of Chinese stock market is mainly displayed through the SSE Composite Index. Therefore, effective prediction of the SSE Composite Index will help investors, enterprises and government agencies to grasp the overall information of the stock market and reduce investment risks in trading. With the development of computer technology, the application of machine learning algorithms to predict stock market volatility has become a hot spot. Among them, BP neural network is the most widely used model, but the gradient descent algorithm used in the model's back propagation has the problem of easily falling into a local minimum. Therefore, the ADAM algorithm to solve this problem was born. This article uses the ADAM optimizer to optimize the BP neural network for SSE Composite Index short-term prediction. The constructed ADAM-BP neural network’s Goodness-of-fit index R2 reached 0.986, which means that the model has good prediction performance. In addition, Compared with the BP neural network without the ADAM optimizer, the error evaluation index of the ADAM-BP neural network is significantly reduced.