A Reinforced-Learning Approach for Calibrating Airlines Itinerary Choice Models with Constrained Demand

摘要

1. Can we develop a model to replicate airlines competition and estimate demand for each itinerary, flight, and market? 1.1 What are itinerary choice models? 1.2 How itinerary choice models are calibrated? 2. What is wrong with the current models? 3. How to model the problem correctly? 4.Model accuracy and Sources of Error. 5. Prediction? 6. Conclusions. 1. Can We develop a Model to Replicate Airlines Competition? Knowing the following: 1. Origin-Destination (OD) demand and fares. 2. Flight schedule of all competing airlines. A logic to generate all competing itineraries from the flight schedule (itinerary builder). A logic to replicate how travelers select itineraries (itinerary choice models). Simulate travelers on the different itineraries using Monte Carlo simulation and keep track of full flights until all demand is loaded.