The Research on the Forecasting of Stock Price Based on the Improved Neural Network

Abstract

In this paper, we reviews the traditional technical analysis and financial models, such as regression and time series model, which often lacks reference value due to that their poor forecasting accuracy and predictive lag. In the age of big data, ANN (artificial neural network) technology has developed rapidly. So, we introduces a variety of classical ANN models and analyzes the traditional use of ANN in financial forecasting. Then, from the perspective of experiential investment, we puts forward the improved neural network and we used it to do empirical analysis with some optimization methods like Dropout. In conclusion, we suggest that the traditional model should be preferred in the short term prediction in the range of acceptable error. On the other hand, the neural network model is recommended if the prediction results and the prediction are required.