Abstract:Predicting edge weights on graphs has various applications, from transportation systems to social networks. This paper describes a Graph Neural Network (GNN) approach for edge weight prediction with guaranteed coverage. We leverage conformal prediction to calibrate the GNN outputs and produce valid prediction intervals. We handle data heteroscedasticity through error reweighting and Conformalized Quantile Regression (CQR). We compare the performance of our method against baseline techniques on real-world transportation datasets. Our approach has better coverage and efficiency than all baselines and showcases robustness and adaptability.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict edge weights in graph - structured data and provide prediction intervals with finite - sample guarantees. Specifically, the author focuses on the load prediction problem in traffic networks. By using graph neural networks (GNN) combined with Conformal Prediction (CP) technology, a new method is proposed to predict the weights of edges in the graph and generate valid prediction intervals. This method is especially suitable for scenarios requiring high reliability, such as traffic flow prediction, where the upper and lower limits of the prediction interval can be interpreted as the upper and lower limits of the predicted traffic flow. ### Main contributions of the paper: 1. **Proposed a method based on Graph Autoencoder (GAE)**: This method can predict the weights of edges in the graph and calibrate the output of GAE through Conformal Prediction technology to generate valid prediction intervals. 2. **Handle data heteroscedasticity**: Through the Error Reweighting and Conformalized Quantile Regression (CQR) methods, the heteroscedasticity problem in the data is handled. 3. **Empirical verification**: The effectiveness of the proposed method is verified on two real - world traffic data sets, and the results show that this method is superior to the baseline methods in terms of coverage and efficiency. ### Core technologies of the paper: - **Graph Neural Network (GNN)**: Used to learn the embedded representations of nodes in the graph. - **Conformal Prediction (CP)**: Used to generate prediction intervals with finite - sample guarantees. - **Conformalized Quantile Regression (CQR)**: Combines CP and Quantile Regression (QR) to handle heteroscedasticity in the data. - **Error Reweighting (ER)**: By assigning covariate - dependent weights to errors, it improves local adaptability and reduces the impact of heteroscedasticity on prediction accuracy. ### Application scenarios of the paper: - **Traffic flow prediction**: In traffic networks, predict the traffic flow of different road sections to support traffic management and optimization. - **Social network analysis**: In social networks, predict the interaction intensity between different users to support recommendation systems and community discovery. - **Biological network analysis**: In biological networks, predict the interaction intensity between different proteins to support drug design and biological research. In conclusion, by combining GNN and CP technologies, this paper proposes a new method to solve the edge - weight prediction problem in graph - structured data and generate prediction intervals with finite - sample guarantees, which has broad application prospects.

Conformal Load Prediction with Transductive Graph Autoencoders

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

Valid Conformal Prediction for Dynamic GNNs

Conformal Inductive Graph Neural Networks

Conformalized Link Prediction on Graph Neural Networks

Uncertainty quantification over graph with conformalized graph neural networks

Conditional Shift-Robust Conformal Prediction for Graph Neural Network

Edge Ranking of Graphs in Transportation Networks using a Graph Neural Network (GNN)

Graph Sparsification for Enhanced Conformal Prediction in Graph Neural Networks

Urban Traffic Forecasting with Integrated Travel Time and Data Availability in a Conformal Graph Neural Network Framework

Similarity-Navigated Conformal Prediction for Graph Neural Networks

Calibrating Graph Neural Networks from a Data-centric Perspective

Improving the interpretability of GNN predictions through conformal-based graph sparsification

BikeNet: Accurate Bike Demand Prediction Using Graph Neural Networks for Station Rebalancing

Refined Edge Usage of Graph Neural Networks for Edge Prediction

Enhancing Trustworthiness of Graph Neural Networks with Rank-Based Conformal Training

Conformal Prediction for Federated Graph Neural Networks with Missing Neighbor Information

Unboxing the graph: Neural Relational Inference for Mobility Prediction

RoCP-GNN: Robust Conformal Prediction for Graph Neural Networks in Node-Classification

GNN at the Edge: Cost-Efficient Graph Neural Network Processing over Distributed Edge Servers

Contrastive Domain-Adaptive Graph Selective Self-Training Network for Cross-Network Edge Classification