Route Intelligence

Predict, classify, and lift profit on every airline route.

A gradient-boosted + neural-network ensemble trained on 16,838 route-months across 49 global carriers and 15 aircraft types, evaluated on a 4,210-flight holdout.

Test flights
4,210
held-out set
Avg profit
$49
per route-month
Profitable
73.7%
3,104 of 4,210
Avg margin
7.1%
test set mean

Model performance

Regression R²
0.700
Operating profit
MAE
$32
Mean abs. error
Classifier accuracy
83.9%
Profitable vs not
AUC
0.899
ROC AUC

Tools