基于离散灰色预测模型与人工神经网络混合智能模型的时尚销售预测

发布时间：2018-03-11 22:22

本文选题：时尚销售预测　切入点：神经网络算法　出处：《计算机应用》2016年12期 　论文类型：期刊论文

【摘要】：时尚销售预测对零售领域十分重要,准确的销售情况预测有助于大幅度提高最终时尚销售利润。针对目前时尚销售预测数据量有限并且数据波动大导致难以进行准确预测的问题,提出了一种结合人工神经网络(ANN)算法和离散灰色预测模型(DGM(1,1))算法的混合智能预测算法。该算法通过关联度分析得到关联度大的影响变量,在利用DGM(1,1)+ANN预测之后,引入二次残差的思想,将实际销售数据与DGM(1,1)+ANN预测结果的残差加入影响变量利用ANN进行第二次残差预测。最后通过真实的时尚销售数据验证算法预测的可行性及准确性。实验结果表明,该算法在时尚销售数据的预测中,预测平均绝对百分误差(MAPE)在25%左右,预测性能优于自回归积分滑动平均模型(ARIMA)、扩展极限学习机(EELM)、DGM(1,1)、DGM(1,1)+ANN算法,相较于以上几种算法平均预测精度大约提高8个百分点。所提混合智能算法可用于时尚销售即时预测,且能够大幅度提高销售的效益。
[Abstract]:Fashion sales forecasting is very important for the retail sector, sales forecast helps to greatly improve the ultimate fashion sales profit. In view of the present fashion sales forecast data and data fluctuation makes it difficult to accurately predict the problems, put forward a combination of artificial neural network (ANN) algorithm and the discrete grey model (DGM (1,1)) prediction algorithm. The hybrid intelligent algorithm this algorithm through correlation analysis of influence degree of correlation of variables in the use of DGM (1,1) +ANN forecast, introduced two residual idea, the actual sales data and DGM (1,1) +ANN prediction residual influence variables using ANN second the residual time prediction. The feasibility and accuracy of the fashion of the real sales data prediction algorithm is verified. The experimental results show that the algorithm in the prediction of fashion sales data, the average absolute forecast Percent error (MAPE) at about 25%, better prediction performance than the autoregressive integrated moving average model (ARIMA), extended the extreme learning machine (EELM), DGM (1,1), DGM (1,1) +ANN algorithm, compared with the above algorithm average prediction accuracy increases by 8 percentage points. The proposed hybrid intelligent algorithm can be used to sales are instant prediction, and can greatly improve the efficiency of sales.

【作者单位】：浙江理工大学科学计算与软件工程实验室;
【分类号】：TP183;N941.5

【相似文献】