АНАЛИЗ АЛГОРИТМОВ МАШИННОГО ОБУЧЕНИЯ ДЛЯ ПРОГНОЗИРОВАНИЯ ЦЕН НА АВИАБИЛЕТЫ

Abstract

The article considers a comparative analysis of several modern machine learning algorithms for predicting airfare prices for the most popular airlines in Kazakhstan. As part of the experiment, all the necessary stages of building machine learning models were completed and possible alternatives were considered. An important element of this kind of task is data preprocessing, which is critical in any machine learning project. This work included a data cleansing process and the use of additional datasets to improve the quality of the results. The set of algorithms considered in this paper was quite wide, and the use of boosting and bagging algorithms proved to be positive. The results of the work obtained using metrics for regression problems can be considered satisfactory and display understandable and readable trends in the data. To improve the results for future studies, it is necessary to use a dataset over a longer period and exclude artificial factors that may influence pricing. It should be considered that the main purchases of air tickets fall on Kazakhstani airlines, which are highly subject to state regulation and the economic situation in the country.