O²RDL-net for joint risk classification and delay forecasting in logistics systems using interaction amplified deep.

Area of Science:

• Operations Research
• Artificial Intelligence
• Logistics Management

Background:

• Transportation and logistics systems are prone to delays and risks from congestion, human factors, and dynamic conditions.
• Current methods often address delay forecasting and risk classification in isolation, missing complex temporal and interaction patterns.

Purpose of the Study:

• To develop a unified deep learning framework for joint delay forecasting and risk classification in logistics.
• To enhance the accuracy and reliability of logistics planning by capturing intricate system dynamics.

Main Methods:

• Proposed O²RDL-Net, a load-aware and interaction-amplified deep learning framework.
• Integrated higher-order operational indicators (Risk Accumulation Index, Delay Sensitivity Indicator, Human-Operational Interaction Score, Logistics System Representation).
• Employed a spatio-temporal attention encoder with risk regime modulation for improved learning.

Main Results:

• O²RDL-Net achieved superior performance over baseline models.
• Delay forecasting: MAE 2.05, RMSE 3.01, R² 0.91.
• Risk classification: Accuracy 92%, Precision 91%, Recall 92%, F1-score 91%.

Conclusions:

• O²RDL-Net offers an accurate, stable, and interpretable solution for real-world logistics risk and delay management.
• The model effectively captures cumulative risk, short-term sensitivity, and system-level interactions.
• Interpretability analysis confirmed the reliance on meaningful, domain-consistent features.