摘要
市场分析和预测已成为企业重要的决策依据和手段。就汽车销量问题提出了一种ARMA模型与RBF神经网络相结合的混合预测方法。采用ARMA模型对汽车销量趋势进行初步线性预测,利用RBF神经网络对线性预测的残差建模,得到非线性预测,两部分预测输出和为总的预测值。该方法既体现了销售量数据间的线性关系, 又揭示了数据内部的非线性特征,克服了单一方法的局限性,提高了预测精度。仿真结构分析表明,该方法预测效果最佳。
A hybrid forecasting method for automobile sale based on autoregressive moving average （ARMA） model and radial basis function neural network （RBFNN） is presented. In the first step, ARMA model is applied to analyzing the linear part of the automobile sale volume. In the second step, an RBFNN is developed to model the residuals from the ARMA model. The final forecast value combines both outputs of ARMA model and RBFNN. Through the proposed method, not only the linear structure of the sale data is exposed but also the nonlinear relationship between data is captured. Simulating experiment results with real data sets indicates that the combined model can be an effective way to improve forecasting accuracy achieved by either of the models used separately.
出处
《天津大学学报（社会科学版）》
2006年第3期175-178,共4页
Journal of Tianjin University：Social Sciences
