# The GECo algorithm for Graph Neural Networks Explanation

Source: [The GECo algorithm for Graph Neural Networks Explanation](<https://www.sota2.com/research/paper/68fa7b451e209131cea876b8>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-25T05:00:04.213Z

Authors: Salvatore Calderaro, Domenico Amato, Giosu\\\`e Lo Bosco, Riccardo Rizzo, Filippo Vella

Published: 2024-11-18T09:08:30.000Z

## Abstract

Graph Neural Networks (GNNs) are powerful models that can manage complex data sources and their interconnection links. One of GNNs' main drawbacks is their lack of interpretability, which limits their application in sensitive fields. In this paper, we introduce a new methodology involving graph communities to address the interpretability of graph classification problems. The proposed method, called GECo, exploits the idea that if a community is a subset of graph nodes densely connected, this property should play a role in graph classification. This is reasonable, especially if we consider the message-passing mechanism, which is the basic mechanism of GNNs. GECo analyzes the contribution to the classification result of the communities in the graph, building a mask that highlights graph-relevant structures. GECo is tested for Graph Convolutional Networks on six artificial and four real-world graph datasets and is compared to the main explainability methods such as PGMExplainer, PGExplainer, GNNExplainer, and SubgraphX using four different metrics. The obtained results outperform the other methods for artificial graph datasets and most real-world datasets.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [GNN Explanation on ba\_cycle\_wheel](<https://www.sota2.com/research/sota/gnn-explanation-on-ba_cycle_wheel.md>): GNN Explanation; ba\_cycle\_wheel; Fid+: 0.866.

- [GNN Explanation on ba\_house\_cycle](<https://www.sota2.com/research/sota/gnn-explanation-on-ba_house_cycle.md>): GNN Explanation; ba\_house\_cycle; Fid+: 0.929.

- [GNN Explanation on er\_house\_cycle](<https://www.sota2.com/research/sota/gnn-explanation-on-er_house_cycle.md>): GNN Explanation; er\_house\_cycle; Fid+: 0.791.

- [GNN Explanation on ba\_cycle\_wheel\_grid](<https://www.sota2.com/research/sota/gnn-explanation-on-ba_cycle_wheel_grid.md>): GNN Explanation; ba\_cycle\_wheel\_grid; Fid+: 0.887.

- [GNN Explanation on er\_cycle\_wheel\_grid](<https://www.sota2.com/research/sota/gnn-explanation-on-er_cycle_wheel_grid.md>): GNN Explanation; er\_cycle\_wheel\_grid; Fid+: 0.915.

- [Graph Explanation on Fluoride-Carbonyl](<https://www.sota2.com/research/sota/graph-explanation-on-fluoride-carbonyl.md>): Graph Explanation; Fluoride Carbonyl; Fidelity+: 0.615.

- [Graph Explanation on Benzene](<https://www.sota2.com/research/sota/graph-explanation-on-benzene.md>): Graph Explanation; BENZENE; FID+: 0.71.

- [Graph Explanation on Alkane-Carbonyl](<https://www.sota2.com/research/sota/graph-explanation-on-alkane-carbonyl.md>): Graph Explanation; Alkane Carbonyl; Fid+: 0.575.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2411.11391>)

[PDF](<https://arxiv.org/pdf/2411.11391.pdf>)
