Flight Passenger-Count Prediction

Good planning depends on knowing how many passengers a given flight is likely to carry, and estimating that from experience alone is unreliable. Get it wrong and the airline either wastes capacity or turns demand away - both expensive at scale.

Turkish Airlines needed a data-driven passenger-demand estimate derived from the characteristics of each flight, rather than a rough manual guess.

I built a regression model using Python and scikit-learn that predicts flight passenger counts from a range of flight features. I prepared and engineered the input features, trained and evaluated regression models, and tuned the pipeline so its predictions generalised to new flights.

The result gives planning teams a quantitative, data-driven demand estimate to work from - a concrete passenger-count prediction per flight instead of an experience-based guess.