extending beyond simple digital reproductions by capturing the spatial, physical, and operational characteristics of structures. Unfortunately, these representations are often complex in nature and difficult to inspect, analyze, and understand which can lead to errors and omissions during model construction. This research aims to leverage graph machine learning systems, utilizing learned datasets, to detect and rectify these issues, improving model quality and minimizing costly mistakes. To illustrate the application of graph neural networks in this domain, this paper applied a graph-based geometric and topological editor coupled with a graph neural network to a real-world dataset of residential building complexes. The developed workflow operates by converting traditional architectural floor plans into graph-structured data, enabling precise node classification predictions. The paper details the overall workflow, data preparation and conversion, hyperparameter optimization and experimental results.
| Authors | Wassim Jabi and Yang Li |
|---|---|
| Year | 2024 |
| Venue | Proceedings of the International Conference on Education and Research in Computer Aided Architectural Design in Europe |
| DOI | 10.52842/conf.ecaade.2024.1.675 |
| Source Database | Seed (CumInCAD) |
| Bridge-to-GNN | Category A |
| GNN Architecture | GraphSAGE (heuristic — verify) |
| Graph Encoding | BIM Element Topology Graph (inferred from title) |
| AEC Task | Element Classification (inferred from title) |
| Cohort | Mature Applications (2024–2026) |
| Dataset Size | 167 samples (heuristic — verify) |
|---|---|
| Implementation Framework | DGL, PyTorch, NetworkX, Topologic (heuristic — verify) |
| Key Hyperparameters | 5 layers; 20 epochs; batch size 16; lr 0.001; optimizer=Adam optimizer; 64 hidden units (heuristic — verify) |
| Primary Metric | Accuracy (heuristic — verify) |
| Primary Value | 95% (heuristic — verify) |
| Key Finding | the concept of Graph Convolutional Networks (GCNs), which are a type of graph neural network designed for semisupervised learning tasks on graph-structured data (heuristic — verify) |
| Quality Assessment | 3/6 · rigor: Medium · Code/Data: ✓Baseline: ✓Reproducible: ✓ (heuristic — verify) |
| GNN architectures detected | GraphSAGE / SAGEConv · GIN · Knowledge Graph Embedding (TransE/RotatE/DistMult) · GCN |
|---|---|
| Frameworks / libraries | Deep Graph Library (DGL) · PyTorch · NetworkX · Topologic / Topologicpy |
| Hyperparameters (regex-detected) | layers=5, 4, 3 · epochs=20, 50 · batch_size=16 · learning_rate=0.001 · hidden_units=64, 69, 17, 86 · optimizer=Adam optimizer |
| Dataset stats found | 167 samples |
| Reported metrics + values | Accuracy: 95% · Accuracy: 94.74% |
| Quality (heuristic, 6-flag) | 2/6 · rigor: Low Code/Data: —Benchmark: —Baseline: ✓CV/Split: —Ablation: —Reproducible: ✓ |
Part of the GML/GNN in AEC Systematic Review (PRISMA 2020) — 112 papers, 2020–2026