Research On Demand Forecasting Model Of Supply Chain Based On Data Mining

In the cross-border e-commerce industry chain,the supply chain demand forecast needs to predict the future demand of each commodity in each warehouse,so that the goods can be prepared in advance in warehouses in various markets around the world,which can effectively reduce logistics time and greatly improve the user experience..This article takes supply chain demand forecasting as the research object,researches a variety of demand forecasting algorithms,and proposes three innovative points for the deficiencies such as the detection of outliers,vectorized representation of product information,and multi-step forecasting.Improved the accuracy of demand forecasting.First,in terms of data processing,a linear regression method based on Huber Loss is proposed to identify outliers in historical sales.Compared with the traditional outlier detection algorithm based on statistics,the outlier detection effect of this method is significantly improved.Secondly,in terms of feature extraction,a product Embedding vector representation based on Pearson correlation coefficient is proposed.In the large-scale product demand forecasting problem,the information representation of products has always been a problem.The product Embedding vector representation method based on Pearson correlation coefficient can learn the low-dimensional vector representation of product information,and it can also reflect the correlation between product sales and Competitive.Finally,optimization of demand forecasting methods at multiple time steps.This paper proposes a deep learning algorithm based on Seq2 Seq + Attention to solve end-to-end time series multi-step prediction,and realizes multi-step prediction in one modeling.The above method is used for experiments on real historical sales data of an e-commerce platform.Experimental results show that the method proposed in this paper has significantly improved the accuracy of prediction compared with traditional time series prediction models and machine learning regression models.