Retail Forecasting Study Based on EEMD-SSA- LSTM Model

Focusing on the problems of multidimensional feature ranking, influencing factor analysis and accurate prediction, this study proposes a data prediction system based on the EEMD-SSA-LSTM model, aiming at the application to the field of new retail merchandise. Firstly, by calculating the cumulative data of single-colour commodities (SKCs) in a specific time period, and combining the entropy-weighted TOPSIS method to rank the comprehensive ability of each SKC, the importance of key features in the comprehensive evaluation is revealed. Secondly, the correlation between multidimensional features, including product features, time factors and external environment, was analysed using Spearman's correlation coefficient, and the impact of these features on sales data was explored. Finally, the EEMD-SSA-LSTM model is applied to predict the time series data with multi-feature inputs, which significantly improves the accuracy and robustness of prediction. This study not only provides advanced algorithmic support for the comprehensive evaluation of features, but also provides a powerful computational tool and practical reference for the analysis and prediction of complex time series data.