In our ongoing commitment to providing the most accurate and comprehensive air travel insights on the market, ForwardKeys is delighted to introduce a groundbreaking innovation that revolutionises airline capacity forecasting — something no one else is currently doing.

Our Data Science team has developed the only solution on the market that produces detailed capacity forecasts extending up to 12 months into the future, broken down by route and flight level. It has been designed to swiftly respond to the multifaceted changes in actual airline performance versus scheduling, reflecting the dynamic responses of carriers to evolving factors.

“The main challenge for travel-dependent industries in relying on scheduled air capacity lies in its volatility. When airlines publish schedules a year in advance, they cannot be certain of the fleet and seat capacity they will use versus the operational reality or demand on that route. The difference between scheduled air capacity (published by the airlines) and actual flown air capacity can vary by up to eight times for specific dates and routes. Factors such as slot allocations, airport limitations, leasing strategies, bilateral agreements on traffic rights, and force majeure events affect the actual picture.”

Daria Kalenskaya
Head of Product & Innovation
forwardkeys-2x-color

In response to these challenges, we have designed an innovative solution comprising three fundamental algorithms:

A sophisticated machine learning model that predicts the likelihood of a scheduled flight being de-scheduled or cancelled.

A complementary machine learning model that forecasts total seats for a given route and flight up to a year in advance.

An algorithm that intelligently combines the outputs of the machine learning models to generate the final, comprehensive Air Capacity Forecast.

The scale of the necessary modelling is significant, with over 110,000 machine learning micro-models being retrained each week — allowing accurate forecasting at the individual flight level. Our continuously improving model treats every route independently and applies hyper-parametrisation to optimise accuracy.

This enables our forecasts to adapt to sudden changes in the aviation landscape, such as changes in flight frequency, operational fleet constraints, commercial adjustments to seat volumes, or even force majeure events like civil unrest, conflicts, or natural disasters. The pioneering predictions from ForwardKeys automatically adjust based on multiple inputs, ensuring that our clients always have access to the most up-to-date and accurate capacity forecasts.

Want the same? Analyse smarter with

Unlock new insights for your business

LEARN MORE

While the legacy of the COVID-19 pandemic presented significant challenges to machine learning models, with 2023 forecasts based on 2020 training data initially proving too low, our team quickly adapted and overcame these anomalies.

We measure the accuracy of our forecasts using the Average Percentage Error (APE), comparing our predictions to actual air seats flown data — and now achieve over 98.7% accuracy on some of the routes, with improvements week on week.

“Looking ahead, we are committed to further refining the accuracy of our capacity forecasting algorithms and exploring opportunities to integrate them with other cutting-edge machine learning models. Our ultimate goal is to provide a truly comprehensive, end-to-end forecast of global airline capacity that empowers our clients to make informed, data-driven decisions.”

Fernando Alcina
Data Science Lead
forwardkeys-2x-color
CONTACT US

Share this, choose your platform!

Categories

Related analysis

2024-07-29T11:57:53+02:0029/07/2024|All, Product|
Go to Top