E-commerce Retailer Demand Forecasting Using Time Series Data:A Two-Stage Temporal Feature Extraction Approach with GBRT Model
Yue Gu, Mingsong Sun · 2023
Amidst the digital economy surge, the e-commerce retail sector experiences unprecedented growth. Yet, merchants grapple with a pivotal challenge: demand forecasting. Existing methods suffer from significant errors, limited dimensions, and smaller solution sizes, hindering precise warehouse demand resolution. To address this, our paper employs a two-stage data feature extraction process, leveraging the GBRT model for forecasting. Initial analysis screens historical data, focusing on key information—merchants, warehouses, and commodities. Employing basic statistical methods and the ARIMA model enhances time-series feature expression, culminating in a regression prediction using the gradient boosting regression tree model. Results affirm the superior predictive efficacy of our two-stage temporal feature extraction model compared to traditional approaches.