Adversarial Multi-Architecture Delay Prediction for Transportation Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional delay prediction methods in transportation networks are inaccurate and network-specific, failing to effectively utilize multimodal operational data for holistic delay prediction, especially in dynamic environments.
Innovation Solution
An adversarial multi-architecture based method that extracts spatial, temporal, and spatiotemporal features using a trained regression model with a critic and regressor network, selecting the architecture with minimum Mean Absolute Error (MAE) for accurate delay prediction in scheduled transportation networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional mathematical models are used for delay prediction, then the system is simple to implement, but the prediction accuracy is low
Solution Approach 1:
The patent transforms the delay prediction problem by changing the modeling approach from conventional mathematical models to deep learning architectures. Multiple regression architectures (fully connected, convolutional, recurrent neural networks) are employed with different parameter configurations to capture complex spatiotemporal patterns in transportation data, significantly improving prediction accuracy despite increased complexity
Solution Approach 2:
The patent creates a composite modeling system by integrating multiple types of neural network architectures into an adversarial multi-architecture framework. This composite approach combines the strengths of different architectures (FCNN for spatial features, CNN for local patterns, RNN for temporal sequences) to achieve superior prediction performance that neither individual architecture could achieve alone
2Measurement precision
If conventional ML methods with common features are used, then the model is easy to train, but it fails to predict accurate delay in dynamic networks
Solution Approach 1:
The patent develops a universal delay prediction framework that can adapt to different transportation networks (railway, bus, metro) through multi-architecture regression models. The system processes various input features (spatial, temporal, spatiotemporal) and adjusts to different network characteristics, making it versatile across multiple transportation modes and dynamic network conditions
Solution Approach 2:
The patent implements dynamic adaptability by training multiple regression architectures simultaneously and selecting the best-performing model for each prediction task. The system dynamically adjusts to changing network conditions and can be applied to different transportation networks without retraining the entire system, enhancing both accuracy and versatility
3Adaptability or versatility
If network-specific models are used, then the model can be trained on available data, but it cannot be implemented for any dynamic network
Solution Approach 1:
The patent creates a universal delay prediction system that works across different transportation networks by employing multiple regression architectures that can be trained on network-specific data and then applied to various dynamic networks. The adversarial multi-architecture framework enables the system to generalize across different transportation modes while maintaining high prediction accuracy
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
The present disclosure predicts a delay associated with a vehicle. Conventional methods are mainly mathematical based and machine learning based networks are not predicting delay accurately. Initially, the present disclosure Initially, the system receives a user query comprising an expected delay of a target vehicle in at least one target station. Further, a real time data associated with the user query in a predefined horizon is obtained. Further, a spatial feature vector, a temporal feature vector and spatiotemporal features are extracted based on the real time data using a feature extraction technique. Finally, the expected is predicted based on the plurality of features using a trained adversarial regression model, wherein the trained adversarial regression model comprises a critic network and a regressor network. The regressor network is trained with a plurality of architectures and a best architecture with minimum Mean Absolute Error (MAE) is selected for delay prediction.