Research On The Prediction Of Stock Based On Optimization Algorithm

With the rapid development of economic and the improvement of comprehensive national strength, equities gradually become an important part of life, and predicting the stock price has become the focus of investors care and research.The stock market is an complex giant system, which is influenced by many factors and uncertainties combined effect, its price fluctuations tend to exhibit strong nonlinear characteristics. The research on stock price prediction is an important research topic.This paper analysis the key issues of stock prediction, combined with technical indicators and the operating characteristics of the stock. Focusing on several methods of prediction on stock price. And multiple of models are proposed to predict the price of stock. Overcome the limitations of single model, and effectively build-up more useful information.The combination forecast can improve the forecast accuracy greatly. Achieve some results in the following areas.First, using a least squares fit stock price based on the analysis Stock Technical indicators and common methods and forecast the stock price of next specific period, combined with the advantages and disadvantages of the simulation results.Second, this paper adopts Gray forecasting method to predict stock price.By analyzing the stock characteristics, combined with the characteristics of gray prediction method, gray prediction algorithm deduced iterative formula for predicting the stock and provides an effective method. The advantages and disadvantages of this method are analyzed combined with its operation.Third, this paper adopts the Neural network to predict the stock. Firstly, BP neural network is adopted to predict the stock, discusses predicting way and gives the simulation results. To overcome the speed and accuracy problem of BP neural network, proposes the dynamic NNARX network to predict. The improved L-M algorithm trains the neural network weights and indirectly determine the number of hidden layer neurons of NNARX. By comparing the theoretical with simulation results analysis, the NNARX network has better prediction results and save training time and frequency.Fourth, this paper proposes the combination forecasting of the stock prices. It is difficult to predict completely only by a predictive model or a certain kind of prediction method because of the uncertainty of random stock price.The different prediction methods are combined in some guidelines appropriately,come up with a suitable predictive models in a variety of situation.In order to adapt to a variety of stocks and many complex case,using genetic algorithm optimizes the combination of weights, optimized by a large number of samples to find the right combination of parameters and solve the profitable weights which is difficult to determine the problem.At the same time it overcomes the limitations of a single method.