EAGLE: Edge-Aware Graph Learning for Proactive Delivery Delay Prediction in Smart Logistics Networks
Authors: Zhiming Xue, Menghao Huo, Yujue Wang
Organizations: College of Engineering, Northeastern University Boston, USA · School of Engineering, Santa Clara University Santa Clara, USA · Department of Chemistry and Chemical Biology, University of New Mexico Albuquerque, USA
Modern logistics networks generate rich operational data streams at every warehouse node and transportation lane -- from order timestamps and routing records to shipping manifests -- yet predicting delivery delays remains predominantly reactive. Existing predictive approaches typically treat this problem either as a tabular classification task, ignoring network topology, or as a time-series anomaly detection task, overlooking the spatial dependencies of the supply chain graph. To bridge this gap, we propose a hybrid deep learning framework for proactive supply chain risk management. The proposed method jointly models temporal order-flow dynamics via a lightweight Transformer patch encoder and inter-hub relational dependencies through an Edge-Aware Graph Attention Network (E-GAT), optimized via a multi-task learning objective. Evaluated on the real-world DataCo Smart Supply Chain dataset, our framework achieves consistent improvements over baseline methods, yielding an F1-score of 0.8762 and an AUC-ROC of 0.9773. Across four independent random seeds, the framework exhibits a cross-seed F1 standard deviation of only 0.0089 -- a 3.8 times improvement over the best ablated variant -- achieving the strongest balance of predictive accuracy and training stability among all evaluated models.
Figures & tables
Figure 1. EAGLE three-module architecture pipeline. (1) PatchTST-Lite: Each node’s T=14 time-step feature sequence is tokenised into Np=2 non-overlapping patches of length p=7 , encoded by a 2-layer Transformer (2 attention heads, dmodel=64 ), and mean-pooled to a fixed temporal embedding ht∈R64 . (2) E-GAT: A 2-layer Edge-Aware GAT (4 heads per layer) aggregates neighbourhood information, conditioning attention coefficients jointly on node embeddings and static edge features Euv (historical transit statistics, shipping mode distributions). (3) DualHead: Two independent MLP heads produce the SLA-violation probability p^v∈[0,1] (sigmoid) and expected delay magnitude d^v∈R+ (Softplus), trained end-to-end via a 0.7×BCE+0.3×Huber multi-task loss. EAGLE three-module architecture pipeline.
Method
F1 (Macro)
AUC-ROC
XGBoost
0.6379±0.0074
0.7455±0.0020
Random Forest
0.6052±0.0018
0.7249±0.0032
LSTM
0.8095±0.0035
0.9679±0.0013
GAT
0.6142±0.0789
0.6872±0.0224
EAGLE (Ours)
0.8762±0.0089
0.9773±0.0037
Table 1. Performance comparison on the DataCo test set (mean ± std, 4 seeds). All methods are evaluated at the node-window level with 14-day sliding window snapshots and an identical relative label definition. For tabular models (XGBoost, RF), predicted probabilities at the order level are averaged within each (node, label window) pair.
Figure 2. EAGLE training dynamics across 4 random seeds. (a) Training loss converges smoothly to ≈0.186 by epoch 40. (b) Validation AUC-ROC stabilises above 0.97 after 15 epochs. Stars mark the best-checkpoint epoch per seed. EAGLE training dynamics across 4 random seeds.
Variant
F1 (Macro)
AUC-ROC
MAE (Days)
A1: No Temporal
0.6851±0.0589
0.8605±0.0698
0.1801
A2: No Edge Features
0.8027±0.0338
0.9540±0.0462
0.0269
A3: Single-Task
0.8230±0.0341
0.9738±0.0120
0.0731†
EAGLE (Full)
0.8762±0.0089
0.9773±0.0037
0.0185
† A3 has no regression head; MAE equals the zero-prediction baseline.
Table 2. Ablation Study Results, and MAE (Days) for the regression sub-task.
Figure 3. Supply chain risk graph with per-node risk scores aggregated across all test snapshots. Node color and size encode the normalized risk score (darker red = higher risk); edges represent active shipping lanes. High-risk hubs (Nodes 10, 22, 17) cluster at the network core, while peripheral low-volume nodes maintain consistently low risk. Supply chain risk graph showing per-node normalized risk scores. Node 10 at the network core has the highest risk score of 1.00, shown in red.
Feature
Type
Source
Time Scope
Future Info?
Justification
order_vol
Node
count(orders)
[t,t+14)
No
Feature window only
mean_scheduled_transit
Node
mean(scheduled_days)
[t,t+14)
No
Known at order placement
std_scheduled_transit
Node
std(scheduled_days)
[t,t+14)
No
Known at order placement
mean_discount_rate
Node
mean(discount)
[t,t+14)
No
Known at order placement
prev_delay_days
Node
mean(actual_delay)
[t,t+14)
No
Past-realised outcomes only
transit_mean
Edge
mean(scheduled_days)
Global static
No
Scheduled (not actual) transit; stable over time
Table 3. Feature audit. Temporal bounds and source for features are recorded to ensure that outcome information does not leak into the features.
Disruptions at critical logistics nodes pose severe risks to global supply chains, yet existing risk prediction systems typically prioritize forecasting accuracy without providing operationally interpretable early warnings. This paper proposes an evidence-grounded framework that jointly performs supply chain bottleneck prediction and faithful natural-language risk explanation by coupling a Temporal Graph Attention Network (TGAT) with a structured large language model (LLM) reasoning module. Using maritime hubs as a representative case study for global supply chain nodes, daily spatial graphs are constructed from Automatic Identification System (AIS) broadcasts, where inter-node interactions are modeled through attention-based message passing. The TGAT predictor captures spatiotemporal risk dynamics, while model-internal evidence -- including feature z-scores and attention-derived neighbor influence -- is transformed into structured prompts that constrain LLM reasoning to verifiable model outputs. To evaluate explanatory reliability, we introduce a directional-consistency validation protocol that quantitatively measures agreement between generated risk narratives and underlying statistical evidence. Experiments on six months of real-world logistics data demonstrate that the proposed framework outperforms baseline models, achieving a test AUC of 0.761, AP of 0.344, and recall of 0.504 under a strict chronological split while producing early warning explanations with 99.6% directional consistency. Results show that grounding LLM generation in graph-model evidence enables interpretable and auditable risk reporting without sacrificing predictive performance. The framework provides a practical pathway toward operationally deployable explainable AI for supply chain risk early warning and resilience management.
Zhiming Xue, Yujue Wang, Menghao Huo
College of Engineering, Northeastern University, Boston, USA · Department of Chemistry and Chemical Biology, University of New Mexico, Albuquerque, USA · School of Engineering, Santa Clara University, Santa Clara, USA
Accurate delivery delay prediction is critical for maintaining operational efficiency and customer satisfaction across modern supply chains. Yet the increasing complexity of logistics networks, spanning multimodal transportation, cross-country routing, and pronounced regional variability, makes this prediction task inherently challenging. This paper introduces a multi-task deep learning model for delivery delay duration prediction in the presence of significant imbalanced data, where delayed shipments are rare but operationally consequential. The model embeds high-dimensional shipment features with dedicated embedding layers for tabular data, and then uses a classification-then-regression strategy to predict the delivery delay duration for on-time and delayed shipments. Unlike sequential pipelines, this approach enables end-to-end training, improves the detection of delayed cases, and supports probabilistic forecasting for uncertainty-aware decision making. The proposed approach is evaluated on a large-scale real-world dataset from an industrial partner, comprising more than 10 million historical shipment records across four major source locations with distinct regional characteristics. The proposed model is compared with traditional machine learning methods. Experimental results show that the proposed method achieves a mean absolute error of 0.67-0.91 days for delayed-shipment predictions, outperforming single-step tree-based regression baselines by 41-64% and two-step classify-then-regress tree-based models by 15-35%. These gains demonstrate the effectiveness of the proposed model in operational delivery delay forecasting under highly imbalanced and heterogeneous conditions.
Stefan Faulkner, Reza Zandehshahvar, Vahid Eghbal Akhlaghi +3
NSF Artificial Intelligence Institute for Advances in Optimization, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA, USA · Kinaxis
The purpose of this research is to find data and methods using machine learning and deep learning to correctly predict the estimated travel time for transportation and logistics in a supply chain system. The supply chain ecosystem is very complex and heavily relies on the transportation and logistics of raw materials and finished goods. Accurate travel time estimation is critical because it helps supply chain members to improve logistics consistency and performance. This helps in planning, demand forecasting, lead time management and assembly planning. The logistics on the delivery side of the customer also plays a crucial role in customer satisfaction and voice of customer. With the collection of huge historical data and using novel techniques, the research builds an accurate model to predict travel time of inventory.