The purposes are to improve the accuracy of inventory demand forecast, balance the indexes of enterprises, and reduce the costs of human, material and financial resources of enterprises and suppliers, thus reducing the supply chain costs and meeting the actual needs of enterprises. In terms of training a large amount of data, deep learning is better than traditional machine learning. The sales demand time series data and the previous material demand time series data are input and trained by back propagation (BP) neural network, and then the material demand value is output. Therefore, the historical data of sales demand forecast and material information are input, and the model is established by BP neural network, which not only takes into account the decisive factor of sales demand forecast, but also considers the material consumption, achieving more accurate forecast. The material demand budget of enterprises is analyzed and a material forecast demand model based on deep learning algorithm is proposed. The model uses a neural network to input the sales demand forecast data, material inventory information and material attribute information into the model, and then the model is trained by the training set in accordance with the error back propagation algorithm. Finally, the training effect of the model is tested by the test set. The results show that when the independent variables include sales demand forecast, material consumption forecast and material attribute information, the forecast error of both models is lower and the effect is better, compared with the material consumption data only as an independent variable. The forecast method based on neural network proposed increases the lead time of the forecast, give the supplier a longer time to prepare goods, and reduce the shortage or surplus of supply caused by the short lead time. Therefore, the material demand forecast model based on convolution neural network (CNN) algorithm provides an important reference for the enterprises, helps them improve their work efficiency and promotes the development of enterprises. This model achieves a great improvement on the accuracy of material demand forecast, and has a certain guiding significance in relevant theory and practice.

目的是提高库存需求预测的准确性，平衡企业各项指标，降低企业和供应商的人力、物力、财力成本，从而降低供应链成本，满足企业的实际需求。在训练大量数据方面，深度学习优于传统机器学习。通过反向传播（BP）神经网络输入和训练销售需求时间序列数据和之前的物料需求时间序列数据，然后输出物料需求值。因此，输入销售需求预测的历史数据和物料信息，通过BP神经网络建立模型，既考虑了销售需求预测的决定性因素，又考虑了物料消耗，实现更准确的预测。分析了企业的物料需求预算，提出了一种基于深度学习算法的物料需求预测模型。该模型使用神经网络将销售需求预测数据、物料库存信息和物料属性信息输入到模型中，然后由训练集按照误差反向传播算法对模型进行训练。最后通过测试集测试模型的训练效果。结果表明，当自变量包括销售需求预测、材料消耗预测和材料属性信息时，与仅作为自变量的材料消耗数据相比，两种模型的预测误差都较小，效果较好。提出的基于神经网络的预测方法增加了预测的提前期，给了供应商更长的备货时间，减少了由于提前期短造成的供应短缺或过剩。因此，基于卷积神经网络（CNN）算法的物料需求预测模型为企业提供了重要参考，帮助企业提高工作效率，促进企业发展。该模型在材料需求预测的准确性上取得了很大的提高，对相关理论和实践具有一定的指导意义。基于卷积神经网络（CNN）算法的物料需求预测模型为企业提供了重要参考，帮助企业提高工作效率，促进企业发展。该模型在材料需求预测的准确性上取得了很大的提高，对相关理论和实践具有一定的指导意义。基于卷积神经网络（CNN）算法的物料需求预测模型为企业提供了重要参考，帮助企业提高工作效率，促进企业发展。该模型在材料需求预测的准确性上取得了很大的提高，对相关理论和实践具有一定的指导意义。