Deep Learning Combinatorial Models for Intelligent Supply Chain Demand Forecasting.

Area of Science:

• Automotive Industry
• Supply Chain Management
• Data Science

Background:

• Growing demand for new energy vehicles (NEVs) driven by environmental concerns.
• The need for accurate demand forecasting to support NEV enterprise decision-making and industry development.
• Intelligent supply chain perspective is essential for optimizing NEV market strategies.

Purpose of the Study:

• To explore NEV demand forecasting within an intelligent supply chain framework.
• To propose and evaluate an innovative combined forecasting model.
• To enhance prediction accuracy and performance for NEV market planning.

Main Methods:

• Development of a novel SARIMA-LSTM-BP combination model for demand prediction.
• Comparative analysis against traditional econometric and deep learning models (Random Forest, SVR, LSTM, BP).
• Evaluation using key performance metrics: Root Mean Square Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE).

Main Results:

• The SARIMA-LSTM-BP model achieved lower RMSE (2.757), MSE (7.603), and MAE (1.912) compared to individual models.
• Demonstrated superior forecasting accuracy and performance.
• Outperformed established forecasting techniques in predicting NEV demand.

Conclusions:

• The SARIMA-LSTM-BP combination model offers a significant advancement in NEV demand forecasting.
• This hybrid approach provides a more accurate and reliable basis for strategic planning in the NEV industry.
• The findings support the adoption of advanced hybrid models for intelligent supply chain management in the automotive sector.