The supply chain demand forecasting model based on LSTM and multiple clustering techniques

Abstract

This study addresses the optimization of supply chain management in e-commerce platforms through the analysis of historical data related to e-commerce activities and product demand. By processing data and conducting anomaly detection, a combination of linear regression, ARIMA, and LSTM models is employed to analyze time series features, with LSTM selected for predicting the demand of various products across different warehouses for each merchant. K-means clustering is utilized to categorize time series data, identifying distinct demand patterns for different products. For new time series data, DBSCAN density clustering and ARIMA models are applied for prediction. Additionally, considering the impact of promotional events such as Singles' Day on demand, ARIMA models are employed to analyze periodic time series data.