基于二次优化BP神经网络的期货价格预测

Futures Prices Forecasting Based on BP Neural Networks of Quadratic Optimization

针对BP算法存在的不足,结合神经网络、遗传算法和主成分分析的优点,提出基于二次优化BP神经网络的期货价格预测算法.初次优化采用主成分分析法对网络结构进行优化,第二次优化采用自适应遗传算法对网络参数进行优化,将经过二次优化后建立的BP神经网络模型用于期货价格预测.经仿真检验,用新方法建立的模型对期货价格进行预测,在预测的精度和速度方面都优于单纯BP神经网络模型.

Future prices were great significance for the futures of dealer, because the model of BP has many problem. This paper gives a new model which based on quadratic optimization BP Neural Network. Quadratic optimization BP Neural Network is combined with neural network, the genetic algorithm and principal component analysis of their respective advantages. First Llsing Principal Component Analysis to optimize the network structure. Second use genetic algorithm optimize weights and threshold of neural network. After a quadratic optimization then establish the BP neural network models for prediction futures prices. Finally, using futures data validation of the new algorithm in predicting the accuracy and speed of both better than simply BP algorithm.