基于粒子群神经网络的期货价格预测

Futures prices forecasting based on PSO neural network

目前在对中国期货市场进行价格预测时,采用神经网络预测时多用的是BP神经网络,但是BP神经网络存在对初始权阈值敏感、易陷入局部极值和收敛速度慢的问题。因此,为了提高模型效率,提出采用PSO-BP模型预测期货价格。首先运用粒子群算法代替BP神经网络的初始寻优,再用BP算法对优化的网络权阈值进一步精确优化,随后建立了基于粒子群算法的BP神经网络预测模型,并将其应用到中国期货市场的期货价格预测研究中。仿真结果表明,新模型结合了粒子群算法的全局寻优能力和BP神经网络算法的局部搜索优势,有效的防止了网络陷入局部极小值的可能,提高了神经网络模型预测的速度和准确性。

At present when using neural network forecasts the Chinese futures market price, mainly using BP neural network. But BP neural network existed initial weight and threshold sensitive, easy to a local minimum point, the slow speed of convergence and other issues. Therefore in order to improve the model efficiency, proposed using PSO-BP model predictive futures prices. First using the PSO algorithm instead of the initial BP neural network optimized, then using BP algorithm to optimize the network threshold value and the right to further optimize the accuracy, finally building forecasting model which Based on the PSO-BP neural network. Its was applicated to China＇s futures market futures prices forecasts for study. The simulation results show that the new model combines the PSO of the global optimization capabilities and BP neural network of local search algorithm advantage, effective prevention network into a local minimum. Improve the neural network model projections of speed and accuracy.