计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (2): 224-229.

• 信号处理 • 上一篇    下一篇

PSO优化BP神经网络的混沌时间序列预测

卢辉斌1,2,李丹丹1,2,孙海艳3   

  1. 1.燕山大学 信息科学与工程学院,河北 秦皇岛 066004
    2.河北省特种光纤与光纤传感实验室,河北 秦皇岛 066004
    3.河北建材职业技术学院 机电工程系,河北 秦皇岛 066004
  • 出版日期:2015-01-15 发布日期:2015-01-12

Prediction for chaotic time series of optimized BP neural network based on PSO

LU Huibin1,2, LI Dandan1,2, SUN Haiyan3   

  1. 1.College of Information Science and Engineering, Yanshan University, Qinhuangdao, Hebei 066004, China
    2.The Key Laboratory for Special Fiber and Fiber Sensor of Hebei, Qinhuangdao, Hebei 066004, China
    3.Department of Mechanical and Electronic Engineering, Hebei Vocational & Technical College of Building Materials, Qinhuangdao, Hebei 066004, China
  • Online:2015-01-15 Published:2015-01-12

摘要: 针对于BP神经网络预测模型,收敛速度慢,精度较低,容易陷入局部极小值等缺点,提出了一种改进粒子群优化BP神经网络预测模型的算法。在该算法中,粒子群采用改进自适应惯性权重和改进自适应加速因子优化BP神经网络预测模型的初始权值和阈值,然后训练BP神经网络预测模型并预测。将该算法应用到几个典型的混沌时间序列预测。实验结果表明,该算法明显提高BP神经网络预测模型的收敛速度和预测模型的精度,减少陷入局部极小的可能。

关键词: 混沌时间序列, 混沌预测, 反向传播(BP)神经网络, 粒子群算法

Abstract: BP neural network for forecasting has low speed of convergence, low precision and easily falling into the local minimum state. An improved prediction method of optimized BP neural network based on Improved Particle Swarm Optimization algorithm(IPSO) is proposed. The IPSO algorithm adopts modified adaptive inertia weight and adaptive acceleration coefficients to optimize the weights and thresholds of BP neural network. Then BP neural network is trained to search for the optimal solution. This experiment is done with several typical nonlinear systems. The results demonstrate that the improved method has faster convergence speed, higher accuracy and not easily falling into the local minimum state.

Key words: chaotic time series, prediction of chaos, Back Propagation(BP) neural network, particle swarm optimization