This study proposes a Cluster-Based Sales Forecasting (CBSF) model for fast fashion (FF) using linguistic and numerical variables. Data are clustered according to the EM algorithm and sales are predicted using extreme learning machines (ELM). The model employs recent real data from a European online FF brand. Results indicate that ELM yields more accurate forecasts than other typical data mining (DM) techniques when applied to CBSF. It also demonstrates that incorporating relevant linguistic variables into the forecasting system and a greater volume of historical data even if from different families, result in improved forecasting. These evidences confirm the relevance of big data to the FF industry.
展开▼