Stock Forecasting Model Based On Neural Network

Stock market is important for national economic development since it is not only an important platform for enterprise financing,but also an important means for investors to manage their finances.Stock market is an important part of national financial system.However,every coin has two sides,on one hand,it can obtain excess returns,on the other hand,it suffers huge risks.How to reasonably predict the future development of the stock market is an urgent concern to investors.In recent years,the rapid development of the national economy has promoted prosperity of the stock market,and the number of stocks has also increased.Some traditional analyses and prediction methods,such as fundamental analysis and technical analysis,have exposed many shortcomings.More and more investors begin to pay attention to the analyze and processing of stock data by some methods of artificial intelligence.Among them,neural network is the typical representative of artificial intelligence applied in stock market.Based on BP neural network and discriminant constrained Boltzmann machine,two stock forecasting models are constructed to analyze and forecast the future price and trend.In the process of building these two models,this paper has done a lot of research work,which can be divided into the following three parts:First,according to the shortcomings of beetle antennae search algorithm,the corresponding improvement strategies are proposed.Although the initial beetle antennae search algorithm has simple structure and fast operation,it is easy to fall into local solution when dealing with multimodal functions.The improved beetle antennae search algorithm increases the number of beetle appropriately and optimizes the population by simplex method after drawing lessons from the characteristics of the group algorithm.The results of numerical experiments show that the improved beetle antennae search algorithm improves the optimization accuracy greatly compared with the original beetle antennae search algorithm,although the operation time increased.Second,the initial parameters of BP neural network are optimized with the help of the improved beetle antennae search algorithm,and then the parameters are further adjusted with BP algorithm,which is applied to the prediction of future stock prices,and the stock price prediction model is constructed.The real data of six stocks such as Ping An Bank are used in the numerical experiment.The numerical experiment results show that the model constructed in this paper has better prediction accuracy than that predicted by BP neural network.Third,a stock forecasting model based on principal component analysis and discriminant restricted boltzmann machine was constructed.Firstly,the principal component analysis is used to reduce the dimension of several stock factors,so that the complexity of the model can be reduced while maintaining the original data information.Then,the principal component after processing is inputted into the discriminant restricted Boltzmann machine to predict the future trend of the stock.The numerical experiments show that the prediction accuracy of this model reaches more than 60%,which can provide valuable reference for investors.