PREDICTING FLIGHT PRICES WITH MACHINE LEARNING METHODS

Abstract

Machine learning is an important area of Artificial Intelligence technologies. Today, thanks to this technology, machines can be successfully trained for difucult tasks in many different areas. This study focuses on machine learning methods and applications that will provide for learning airline price changes gradually from past flight patterns to predict future prices. Flight prices vary greatly depending on airlines' policies, holidays, student mobility, number of remaining seats, months and days, even hours. Considering this situation along with the competition factor and revenue maximization, it is a very difficult task for airline companies to determine ticket prices under the influence of many factors in the most appropriate way. Especially ensemble learning algorithms and many algorithms such as Support Vector Machines, Random Forest, Gradient Boosting, K-Nearest Neighborhood algorithms are used either alone or together to accomplish this difficult task. In this study, machine learning applications developed for flight price prediction were investigated and these applications were analyzed in detail in terms of the method, data set and application performances used. As a result, it is seen that ensemble machine learning models combining strengths (by compensating weaknesses) of multiple sub-models are more successful in predicting flight prices. With the knowledge and experiences obtained from this study, new models that can be used for flight price prediction by using ensemble learning algorithms have been developed within the scope of the "Flight Prices Predictor" project carried out by the ENUYGUN.COM R&D Research Center.