Research on E-Commerce Demand Forecasting and Inventory Optimization Based On SARIMA Time Series and Linear Regression

In this paper, we start from solving the supplier’s demand forecasting, in order to solve the e-commerce retail merchant’s demand forecasting and inventory optimization problem, we carry out the demand forecasting and inventory optimization, and establish an accurate demand forecasting model to predict the future demand of each warehouse. In this paper, some overall data about merchants, commodities and warehouses are selected. Then merge using merge function in python. Outliers are processed through 3σ principle. The data is transcoded and then analyzed. ARIMA and SARIMA time series forecasting models are established to find the pattern of demand for goods, and the forecasting results are computed by elbow rule and K-means clustering method to arrive at the optimal solution for the number of classifications, which further categorizes the time series of merchants, commodities, and warehouses. Then the sales data of different months are integrated and analyzed descriptively. Meanwhile, the construction of linear regression function is carried out and the linear regression equation is derived. By analyzing the fluctuation law trend, the demand deviation value is derived. After that, a forecasting model based on the superposition of commodity trends is established, and the forecasting data are finally obtained. Through research to address demand forecasting in e-commerce retail supply chain management, the integration and utilization of data is improved, advanced data analysis and mapping analysis techniques and robust clustering methods are used to classify the data and discover the laws, which ensures the accuracy and optimality of the parameters of the subsequent forecasting model, and finally demand forecasting is carried out, and multi-dimensional information such as historical commodity attributes, demand data, and users’ personalized demands are In-depth analysis of multi-dimensional information such as historical product attributes, demand data, users’ personalized needs, etc., establishes an accurate and flexible demand forecasting model.