Downloads · 30 days
5
23% of all-time downloads
alirezaaminzadeh/routelab-solver-selector
routelab-solver-selector is a machine learning model from alirezaaminzadeh. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for routelab. The card lists the license as apache-2.0.
Instance-to-solver routing model that predicts whether RouteFinder, OR-Tools, PyVRP, SCIP, or the Hybrid pipeline will perform best on a given VRP instance.
Downloads · 30 days
5
23% of all-time downloads
All-time downloads
22
Public
Repo size
—
Likes
0
Public
Click a slice to open those files.
.json3.3 KB · 47%
From the Hugging Face model README
Instance-to-solver routing model that predicts whether RouteFinder, OR-Tools, PyVRP, SCIP, or the Hybrid pipeline will perform best on a given VRP instance.
Expert-calibrated ensemble selector (no training required). Uses 13 instance features to rank solvers by predicted gap, runtime, and feasibility probability.
| Feature | Description |
|---|---|
n_customers | Number of customers |
n_vehicles | Fleet size |
capacity_utilization | Total demand / fleet capacity |
customer_clustering | Spatial cluster coefficient |
demand_variance | Demand distribution spread |
time_window_tightness | Time window constraint density |
route_openness | Open vs closed route flag |
backhaul_ratio | Fraction of backhaul customers |
stochastic_factor | Demand/travel uncertainty level |
dynamic_arrival_rate | Dynamic customer arrival rate |
graph_density | Customer density metric |
avg_nearest_neighbor | Mean nearest-neighbor distance |
depot_centrality | Depot centrality in customer cloud |
{
"recommended_solver": "hybrid",
"recommended_family": "hybrid",
"confidence": 0.82,
"rankings": [
{"solver_id": "hybrid", "score": 0.74, "predicted_gap_pct": 0.4},
{"solver_id": "routefinder", "score": 0.68, "predicted_gap_pct": 2.0},
{"solver_id": "ortools", "score": 0.61, "predicted_gap_pct": 0.9}
]
}
| File | Description |
|---|---|
config.json | Model metadata and performance metrics |
selection_weights.json | Calibrated ensemble weights per solver |
Evaluated on 21-instance medium-size benchmark suite:
config.jsonconfig.json