Study On The Prediction Of Stock Price Based On Neural Network Technology

With China's rapid economic growth and the expansion of the financial market,the stock has become an important part of people's economic life.People have tried to study the stock price volatility and master the regular pattern of price fluctuations since the stock market came into being.The BP neural network,as a classic algorithm of the big data prediction,is favored by investors and researchers.However,it also has defects,for example,the BP algorithm itself shows slow convergence speed so it is quite easy to fall into the local minimum point defects,thus leading to the low efficiency of the forecast.By analyzing the basic problems faced by the stock price forecasting and comparing a variety of stock price forecasting methods,this paper discusses the BP neural network algorithm for predicting the stock price method.Genetic algorithm is found to search out the best individual as BP neural network initial weights and thresholds by using trial and error method to calculate the specific number of hidden layer nodes.Then,through the BP algorithm training,the training process and the learning rate are automatically adjusted.Based on genetic algorithm,the BP neural network can play a part not only in its global search of genetic algorithm but also in its efficiency,the combination of the two parts can speed up the computation efficiency and enhance the BP neural network's ability in learning and forecasting the stock.Based on the algorithm idea,this paper constructs a predicting model which using Matlab7.0 as the experimental platform for simulation experiment and selecting 300 index in Shanghai and Shenzhen in 2014 as the experimental object.The model uses historical data for the previous ten days to predict the eleventh day's closing price.Among them,the historical data for the 140 trading days in the first 100 sets of data as the training sample,30 sets of data as the test sample.The single BP neural network predicting model and BP neural network and genetic algorithm are also compared and analyzed from the following three aspects as the convergence speed,stability and predicting results.Experiments show that the proposed predicting method based on genetic neural network's stock has practicability to some extent.