论文部分内容阅读
为实现对于时间序列预测数据的准确预测,提出一种神经元拓展正则极端学习机(NERELM,Neuron-Expanding Regularized Extreme Learning Machine),并研究了其在时间序列预测中的应用.NERELM根据结构风险最小化原理权衡经验风险与结构风险,以逐次拓展隐层神经元的方式自动确定最佳的网络结构,以避免传统神经网络训练过程中需人为确定网络结构的弊端.应用于时间序列的仿真结果表明:NERELM可有效实现对于RELM最佳网络结构的自动确定,具有预测精度高与计算速度快的优点.
In order to achieve the accurate prediction of time-series prediction data, a neuron-Expanding Regularized Extreme Learning Machine (NERELM) is proposed and its application in time series forecasting is studied.NERELM is based on the least structural risk The principle of trade-off weighs on experience risk and structure risk, and automatically determines the best network structure by expanding the hidden neurons one by one in order to avoid the drawbacks that the artificial neural network structure needs to be determined in the traditional neural network training process. The simulation results applied to the time series show : NERELM effectively automates the best network structure for RELM, with the advantages of high prediction accuracy and high computational speed.