基AHP和BP神经网络的服装销售预测模型的研究及应用

发布时间：2018-05-16 13:30

本文选题：层次分析 + BP神经网络　；参考：《浙江工商大学》2014年硕士论文

【摘要】：对于服装行业来讲,企业运营效率的提高主要依托于两个要素,一个是服装销售预测的精度,另一个是供应链的反应速度,即供应链的效率。如果预测的精度高,则即便供应链的反应速度不够快,也能够实现库存与资金的高周转,可见服装销售预测的准确度对于公司上层做决策起到重要的作用。服装产业是一个复杂的、销售渠道多样的、快速变化的时尚产业,因此短周期预测对服装产业更为合适。针对这一特点,本文除常规的季节、周期、节假日外,增加了竞争对手、流行走向等因子,全面提取了影响服装销售的影响因素,从而使销售预测在更大层度上逼近真实的销售量。本文将层次分析法(Analytical Hierarchy Process,简称AHP)与BP (Back Propagation)人工神经网络相结合,提出了一套A-BP销售预测模型,将定性预测算法与定量预测算法紧密的结合了起来。该模型的构建过程是：首先提出了多个影响销售量的因子,然后将这些影响因子通过层次分析建立三级层次结构,通过两两比较确定层次中因子的相对重要性,输出各个因子对销售量的权重,最后将权重值排在前几位的影响因子纳入BP神经网络中进行计算,得出预测销售量。最后,本文实现了一个基于A-BP模型的销售预测系统,通过实验验证了A-BP销售预测模型比BP神经网络模型预测准确更高的结论。
[Abstract]:For the garment industry, the improvement of enterprise operation efficiency mainly depends on two factors, one is the precision of clothing sales prediction, the other is the response speed of supply chain, that is, the efficiency of supply chain. If the prediction accuracy is high, even if the supply chain reaction speed is not fast enough, it can also realize the high turnover of inventory and capital. It is clear that the accuracy of clothing sales prediction plays an important role in the upper layer of the company to make decisions. The garment industry is a complicated, diversified and rapidly changing fashion industry, so short period prediction is more suitable for garment industry. In view of this characteristic, in addition to the normal season, cycle, holiday, this article has added the competition, the popular trend and so on factor, has completely extracted the influence factor which affects the clothing sale, So that the sales forecast in a larger level approach to the real sales volume. In this paper, the Analytical Hierarchy process (Hierarchy) and BP back Propagation (BP) artificial neural network are combined, and a set of A-BP sales forecasting model is proposed, which combines the qualitative prediction algorithm and the quantitative prediction algorithm closely. The construction process of the model is as follows: firstly, several factors affecting sales volume are put forward, then the three-level hierarchical structure is established by hierarchical analysis, and the relative importance of the factors in the hierarchy is determined by comparison. The weight of each factor to the sales volume is outputted. Finally, the influence factors which rank the weight in the first place are incorporated into the BP neural network to calculate the predicted sales volume. Finally, a sales forecasting system based on A-BP model is implemented in this paper. The experimental results show that the A-BP model is more accurate than the BP neural network model.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.52;TP183

【参考文献】