City-Networks

Shanghai

Node Classification PyG dataset page ↗PyG ↗

Node classification on the Shanghai road graph, predicting an accessibility class for every intersection. Street segments are edges and labels depend on long-range network structure. Evaluated by Accuracy.

Dataset details

Task

Shanghai road-network node classification from the City-Networks benchmark. Labels are derived from long-range accessibility statistics on the city graph. The primary catalog metric is Accuracy.

Data

Shanghai road-network node classification from the City-Networks benchmark. Labels are derived from long-range accessibility statistics on the city graph. The cataloged artifact reports 1 graphs.

Features

Road-network intersections are nodes and street segments are edges. The release derives node inputs and accessibility-based labels from the embedded city network; adding coordinates, centralities, or external map attributes changes the feature set.

Splits and evaluation

The catalog records these protocols or variants: Standard split. Evaluation uses Accuracy (higher is better). Exact masks or folds must come from the cited release.

Quirks and pitfalls

The four cities differ in topology, sampling, and label distribution, so they are separate datasets rather than interchangeable folds. Labels are derived from the same graph, creating a risk of target leakage through engineered centrality or accessibility features.

Sources

SOTA Progression

new SOTA matches SOTA at that time other proposed result other baseline result

Leaderboard

#ModelAccuracy ↑Paper
1
GAT
0.7253 ± 0.0023
2
Exphormer
0.7020 ± 0.0040
3
GatedGCN
0.6880 ± 0.0021
4
GraphSAGE
0.6830 ± 0.0050
5
GAT
0.6800 ± 0.0050
6
GPS + Performer
0.6727 ± 0.0017
7
GPS + k-MIP
0.6694 ± 0.0044
8
ChebNet
0.6650 ± 0.0010
9
GPS + BigBird
0.6524 ± 0.0017
10
SGFormer
0.6410 ± 0.0030
11
GINE
0.6335 ± 0.0020
12
GraphGPS
0.6300 ± 0.0050
13
Exphormer
0.6233 ± 0.0016
14
GCN
0.6210 ± 0.0020
15
GCNII
0.6150 ± 0.0040
16
DropEdge
0.6080 ± 0.0040
17
GCN
0.5775 ± 0.0024
18
MLP
0.2840 ± 0.0060
SOTA performance set a new SOTA at the time matched the SOTA at the time