基于季节性RBF神经网络的月度市场需求预测研究

本文提出一种季节性神经网络预测模型,对具有季节性变化的产品月度市场需求进行预测.在Matlab语言环境下,用傅立叶周期分析法得到时间序列的周期长度;借鉴嵌入理论,提出了确定季节性神经网络输入维数的策略;利用计算机程序搜索,确定最优参数;通过合理插值,重构样本集.仿真实验表明,该模型的预测精度明显高于其他几个常用的季节预测模型.

Study on Forecasting the Monthly Market Demand Based on Seasonal RBF Neural Network

To forecast the monthly market demand of some products, the seasonal RBF neural network forecasting model is introduced.On the platform of Matlab software,the seasonal length is calculated by Fourier period analysis method;the method of calculating the input dimension is proposed based on the embedding theory;the optimization parameters are obtained by computer program searching;and the sample set is reconstructed using linear interpolation.The simulated experiment results indicate that the forecasting precision of the seasonal RBF neural network is clearer than the other common seasonal forecasting models.