Airline demand forecasting — archived coursework

Archived 2018 Seattle University IS 5201 group project by Afsar Ali, Matthew Ensslen, and Russell Key under Dr. Misuk Lee. The team evaluated basic additive and multiplicative airline-booking forecasts against a naive baseline. This is academic work, not a production airline system.

View Afsar Ali’s portfolio index.

Scope, team, and data

• The project summary identifies the team as Afsar Ali, Matthew Ensslen, and Russell Key.
• Individual contribution shares are not documented.
• The training file contains 5,185 rows and three columns; the validation file contains 203 rows and five columns.
• The original data provider and redistribution rights are not documented. No license is granted by this repository.

What the project demonstrates

• booking-date and departure-date feature engineering;
• daily booking deltas and three-day smoothing;
• day-of-week aggregation;
• additive and multiplicative booking forecasts;
• validation against a naive forecast.

The archived team summary reports relative-error results of 70.7% for the additive approach and 83.0% for the multiplicative approach, with the additive approach performing better. These are historical reported results; they were not reproduced under a current environment.

Artifacts

• Project summary
• Python source
• Training data
• Validation data

Runtime and reproducibility status

The source is Python 2-era code using pandas and numpy. Python and dependency versions were not recorded, and the repository has no dependency lockfile, tests, continuous-integration workflow, or expected-output fixture. A Python 3.12 syntax check fails on the Python 2 print statement; the code also uses a pandas date API removed from current releases. The code has not been ported or otherwise changed as part of this documentation cleanup.