MCDFN: Supply Chain Demand Forecasting via an Explainable Multi-Channel Data Fusion Network Model

TL;DR

This paper tackles the challenge of accurate and explainable demand forecasting in supply chains by proposing MCDFN, a hybrid multi-channel neural network that fuses CNN, BiLSTM, BiGRU, and multi-layer LSTM branches to capture spatial and temporal patterns in time-series data. It systematically preprocesses a 1826-day retail dataset with cyclic date features and holidays, trains multiple baselines, and demonstrates that MCDFN achieves superior performance across $MSE$, $RMSE$, $MAE$, and $MAPE$, with robust statistical validation. The authors also address interpretability by applying ShapTime and Permutation Feature Importance, revealing the predominance of cyclic temporal features and the impact of holidays. The study provides practical guidance for integrating MCDFN into SCM systems, highlights a favorable balance of accuracy and inference speed, and outlines future work on scalability, real-time deployment, and broader domain applicability.

Abstract

Accurate demand forecasting is crucial for optimizing supply chain management. Traditional methods often fail to capture complex patterns from seasonal variability and special events. Despite advancements in deep learning, interpretable forecasting models remain a challenge. To address this, we introduce the Multi-Channel Data Fusion Network (MCDFN), a hybrid architecture that integrates Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Gated Recurrent Units (GRU) to enhance predictive performance by extracting spatial and temporal features from time series data. Our comparative benchmarking demonstrates that MCDFN outperforms seven other deep-learning models, achieving superior metrics: MSE (23.5738), RMSE (4.8553), MAE (3.9991), and MAPE (20.1575%). Theil's U statistic of 0.1181 (U<1) of MCDFN indicates its superiority over the naive forecasting approach, and a 10-fold cross-validated statistical paired t-test with a p-value of 5% indicated no significant difference between MCDFN's predictions and actual values. We apply explainable AI techniques like ShapTime and Permutation Feature Importance to enhance interpretability. This research advances demand forecasting methodologies and offers practical guidelines for integrating MCDFN into supply chain systems, highlighting future research directions for scalability and user-friendly deployment.

Figures (13)

• Figure 1: Proposed MCDFN methodological framework for supply chain demand forecasting.
• Figure 2: Yearly product consumption pattern (from 2013 to 2017).
• Figure 3: Time-series analysis of sales data and feature engineered cyclic features over the course of a year.
• Figure 4: Violin plot of the normalized features
• Figure 5: Data splitting into training, validation, and test sets