Scalable transit delay prediction at city scale: A systematic approach with multi-resolution feature engineering and deep learning

Keyphrases

• Deep Learning 100%
• Feature Engineering 100%
• City Scale 100%
• Engineering Learning 100%
• Multi-resolution Features 100%
• Transit Delay 100%
• Delay Prediction 100%
• Spatial Regions 66%
• Adaptation 33%
• Network Applications 33%
• Prediction Accuracy 33%
• Spatio-temporal Features 33%
• Model Driven Architecture 33%
• Dimensionality Reduction 33%
• Handcrafted Features 33%
• Transformer Model 33%
• Waiting Time 33%
• Prediction pipeline 33%
• Feature Space 33%
• Compressed Features 33%
• Prediction System 33%
• Temporal Patterns 33%
• Comparable Accuracy 33%
• Compact Architecture 33%
• XGBoost 33%
• Elementary Level 33%
• Route Segment 33%
• Network Reliability 33%
• Clustering Problem 33%
• Validation Analysis 33%
• Scale Prediction 33%
• Dense Urban Areas 33%
• Bus Operation 33%
• XLSTM 33%
• Multilevel Evaluation 33%
• Distributed Training 33%
• Fewer Parameters 33%
• Temporal Dependency 33%
• Urban Bus Transport 33%
• Network Topology Information 33%
• Giant Cluster 33%
• Hybrid Clustering 33%
• Autoformer 33%
• Cluster-aware 33%
• Real-time Operational Control 33%
• Transit Agencies 33%
• Adaptive PCA 33%
• Walk-forward Validation 33%
• Space-making 33%
• Reusable Architecture 33%
• Latency Analysis 33%
• Hierarchical Index 33%
• PatchTST 33%
• Route Type 33%
• Operations Staff 33%
• Overparametrization 33%

Computer Science

• Deep Learning Method 100%
• Long Short-Term Memory Network 100%
• Feature Engineering 100%
• Model Architecture 50%
• Dimensionality Reduction 50%
• Clustering Method 50%
• Network Topology 50%
• Feature Space 50%
• Extreme Gradient Boosting 50%
• Transformer Model 50%
• Operation Staff 50%