Product Sales Prediction with Variational Mode Decomposition and Multi-Scale Feature Extraction

With the rapid development of digital technology and business models, significant changes have occurred in the business environment. Product sales prediction is of great significance for inventory management, supply chain optimization, and market strategy formulation. In this paper, we propose a novel product sales prediction method VMD-PyraFormer based on Variational Mode Decomposition (VMD) and multi-scale time-dependent features. To address the complex nonlinearity and volatility in product sales data, we adopt VMD to decompose the product sales time sequence into multiple modal components and capture various periodic features. Subsequently, PyraFormer is constructed as the feature extraction encoder to extract trend and fluctuation features at different time scales. Comparative experiments were conducted on real product sales dataset. The experimental results show that the proposed method has better prediction accuracy than traditional time series prediction methods. We also conducted ablation studies to further explore the characteristics of VMD and PyraFormer. Comprehensive experimental analysis demonstrated the effectiveness of the proposed method. The proposed sales prediction model is of great significance in assisting enterprise decision-making and improving enterprise competitiveness. It also provides a new solution for theoretical and technological innovation in sales prediction.