CLASSIFICATION TECHNIQUES FOR AIRFARE FORECASTING

BASIC INFORMATION

Ph.D. Student: Marco Antonio Barrón
Advisors: Jose M. Luna, Sebastián Ventura
Defended on: Mach 2023
Keywords: Gene expression, airline fare war, classification, recommender systems
Digital version

DESCRIPTION

This work is focused on the multi-factorial problems that commercial airlines face up, such as the pricing war and the creation of a dynamic discount table through the implemen- tation of evolutionary algorithms and data mining methods. On the one hand, in the airline industry, the Revenue and Pricing teams generally spend a considerable amount of time analysing and interpreting the actions of their competitors. Most of the time the analysts have to use their analytical skills to create ad-hoc methods to interpret or find patterns in the fares. The use of automatic methodologies is key to reducing time and avoiding human errors. This thesis proposes:

• A new methodology to predict, analyze and interpret airline fares which are capable of mimicking manual processes executed by pricing teams. A gene expression programming algorithm is proposed to mimic the manual process carried out by pricing teams by adding new features automatically. The algorithm can explore huge search spaces, which is a daunting process to be done manually as pricing teams do daily. A real scenario was considered in the experimental analysis by considering Air Canada fares in the period December 2019 to January 2020, corresponding to a travel period between December 2019 and April 2020.
• On the other hand, historically, airlines around the globe have used static pricing structures, which are constrained to discrete price points and there is limited segmentation between their guests. Because of these limitations and constraints, the necessity of novel methods to calculate the willingness to pay and identify potential guests whose propensity to book a flight will increase if they receive a discount to improve their sales is huge. Thus, this thesis proposes a novel methodology to identify interesting subgroups whose chance to book a flight increases if they receive an offer discount. This proposal includes a grammatically evolutionary feature selection algorithm to extract the best subgroups by analyzing the booking behaviour of historical passengers. A real case scenario was considered in the experimental analysis using private data from a commercial airline.

FUNDS

• Spanish Ministry of Science and Competitiveness, project PID-2020-115832GB-I00.

PUBLICATIONS ASSOCIATED WITH THE THESIS

INTERNATIONAL JOURNALS

1. Facing up fare war: generating competitive price models with gene expression pro- gramming. Barron, Marco Antonio, Jose Maria Luna, and Sebastian Ventura. IEEE Access, 2022.

INTERNATIONAL CONFERENCES

1. Dynamic Airline Discounts using an Evolutionary Subgroup Discovery Methodology. Barron, Marco Antonio, Jose Maria Luna, and Sebastian Ventura. 2022 IEEE Inter- national Conference on Omni-layer Intelligent Systems (COINS). IEEE Computer Society, 2022.