中小企业肉类产品短期需求预测模型研究

发布时间：2019-03-29 19:49

【摘要】：近年来,随着中国在国际上影响力的提升,我国食品企业所面临的竞争压力逐渐增大。为了更好地参与国际竞争以及应对国内产业结构变化,中小型企业必须要推动自身管理体制的改革。目前我国食品企业普遍面临库存占用过大的问题,已成为制约企业发展的瓶颈,易变质及存储成本高的肉类产品特性促使此库存问题尤为突出。引发此类问题的主要原因是对短期需求预测不精确,因此有必要对中小企业肉类产品的短期需求预测进行研究。鉴于此,本文基于灰色理论、BP (Back Propagation,反向传播)神经网络、SVM (Support Vector Machine,支持向量机)及 GA( Genetic Algorithm,遗传算法)构建肉类产品短期需求预测模型。通过实验对比,选取出精确度较高且拟合度较好的预测模型。首先,选取中小企业肉类产品短期需求的主要影响因素。根据影响因素选取原则,最终选取月度CPI (Consumer Price Index,居民消费价格指数)、产品价格、促销支出成本、季节系数和节假日系数5个主要因素。其次,选择预测方法。分析现有的预测方法,对比传统预测方法和人工智能预测方法,选取预测精确度较高的人工智能预测方法进行研究。并从人工智能预测方法中选取较为常用的BP神经网络和SVM两种预测方法。然后,构建预测模型。本文提出将灰色预测理论同BP神经网络和SVM相结合,避免因样本数据随机性过大导致预测精度降低的问题。为了解决BP神经网络和SVM的初始参数值具有随机性的问题,引入GA分别对两个预测方法进行优化。最后,对比分析实验结果。利用灰色BP神经网络、灰色GA-BP神经网络、灰色SVM及灰色GA-SVM预测模型分别对肉类产品月需求量进行预测,并依据后验差检验法和残差检验法对四个模型进行评价。研究表明,灰色GA-SVM模型对中小型企业肉类产品的月需求量预测精确度较高且拟合度较好。
[Abstract]:In recent years, with the promotion of China's influence in the world, the competition pressure of Chinese food enterprises is increasing gradually. In order to better participate in international competition and cope with the change of domestic industrial structure, small and medium-sized enterprises must promote the reform of their own management system. At present, food enterprises in China are generally faced with the problem of excessive inventory occupation, which has become a bottleneck restricting the development of enterprises. The characteristics of meat products, which are prone to deterioration and high storage costs, make this inventory problem particularly prominent. The main cause of this kind of problem is the imprecise short-term demand forecast, so it is necessary to study the short-term demand forecast of small and medium-sized meat products. In view of this, this paper constructs the short-term demand forecasting model of meat products based on grey theory, BP (Back Propagation, back propagation (, SVM (Support Vector Machine, support vector machine) and GA (Genetic Algorithm, genetic algorithm (GA). Through the comparison of experiments, the prediction model with high accuracy and good fitting degree is selected. First of all, select the main influencing factors of short-term demand for meat products of small and medium-sized enterprises. According to the selection principle of influencing factors, five main factors are selected: monthly CPI (Consumer Price Index, consumer price index (CPI), product price, promotion cost, seasonal coefficient and holiday coefficient. Secondly, the prediction method is selected. The existing forecasting methods are analyzed, compared with the traditional forecasting methods and artificial intelligence forecasting methods, and the artificial intelligence forecasting method with high prediction accuracy is selected to carry on the research. The commonly used BP neural network and SVM prediction method are selected from the artificial intelligence prediction method. Then, the prediction model is constructed. In this paper, the grey prediction theory is combined with BP neural network and SVM to avoid the problem of decreasing the prediction precision due to the large randomness of sample data. In order to solve the problem that the initial parameter values of BP neural network and SVM are random, GA is introduced to optimize the two prediction methods respectively. Finally, the experimental results are compared and analyzed. Grey BP neural network, grey GA-BP neural network, grey SVM and grey GA-SVM forecasting models were used to forecast the monthly demand of meat products, and the four models were evaluated according to the posterior error test and the residual test. The results show that the grey GA-SVM model can predict the demand of meat products in small and medium-sized enterprises with high accuracy and good fitting degree.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F276.3;TP18

【参考文献】