Research on Sales Dynamic Forecasting Method Based on Time Series Analysis in Global Supply Chain Environment

This paper uses time series analysis to forecast sales dynamics in a global supply chain environment. The study selects supermarket chain data with typical time series characteristics and extracts features from it. We combine the delay operator’s difference operation with the ARMA model to construct the ARIMA model, which predicts and analyzes the sales volume of the supermarket chain dataset from 2016 to 2018. To address the issue of time series models being susceptible to nonlinear characteristics and random variables, this paper integrates the ARIMA model and the random forest RF model into a simple weighted average, forming the ARIMA-RF model. The accuracy of the model for sales volume prediction has greatly improved compared to the ARIMA model and RF model, with an accuracy rate as high as 99%. Meanwhile, the ARIMA-RF model has the smallest RMSE value among all the compared models and has the highest prediction accuracy among all the models for the sales volume in the 3rd quarter of 2018.