Computer intelligent quantitative transaction decision model based on GRU network

Abstract

We build a model that gives the best daily trading strategy only based on the day's price data. Firstly, we use the RNN neural network model as the basic model. However, we find that there will be problems such as gradient disappearance or gradient explosion when training the network due to the cumulative rise of information. So we have made improvements to use the GRU (Gated Recurrent Unit) network, which can predict the next day's price. Then, we use the Apriori data mining algorithm to preprocess data and establish a Quantitative transaction decision model. However, the obtained solution is too complex, and we carry out nonlinear fitting into an exponential trading formula. The fitting effect is better than previous results.