Prediction Model Of Stock Index Based On PSO-optimized Chaotic BP Neural Network

With the stock market investment activities become more frequent, urgent need for an effective market forecasting method to help people increase revenue while appropriate risk reduction. Highly complex markets, which have a certain trend in the variation, but by political, economic, psychological and other factors, changes of the market is still elusive. The traditional method of stock index forecasts are based on a large number of statistical method, is to first establish a common feature of the subjective data series model, and then calculated according to the subjective model and predict, forecast accuracy can not meet the actual requirements. Chaotic stock index has obvious characteristics, many scholars have conducted in-depth study chaotic properties, the establishment of a variety of stock index based on chaos theory prediction model. Chaos BP neural network model is a successful forecasting model pre-stock index, but the models are easy to fall into local minima and slow convergence of the shortcomings, thereby increasing chaos BP neural network prediction accuracy is the prediction of stock index an important issue.First, the establishment of a chaos theory-based BP neural network prediction model will be applied to two kinds of typical nonlinear time series (Logistic chaotic system and the Henon chaotic system) and the Shanghai Composite Index measured time series validation and gives the predictive value of different training samples of comparison and analysis of results. To further improve the forecast accuracy of the prediction model, a chaotic particle swarm optimization based on BP neural network improved prediction model. This model uses the number of input and output parameters of time series constructed BP neural network topology, the use of chaotic particle swarm optimization algorithm BP neural network weights and thresholds, will improve the prediction model applied to the Logistic, Henon chaotic time series and the typical measured Shanghai Composite time series validation. The results show that particle swarm optimization chaos after the BP neural network model can effectively compensate for lack of BP neural network to improve the BP neural network easy to fall into local optimum, to a certain extent, to achieve the purpose of improving the performance of the algorithm. The model of two kinds of typical nonlinear chaotic time series and measured the Shanghai Composite Index has better capacity and higher nonlinear fitting prediction accuracy.