Year: 2024
Type: article
Abstract: The article presents the results of using machine learning methods to study data from a special questionnaire that considers the general characteristics of flights, the characteristics of passengers and their opinions on various aspects of the flight. The objective of the study is to identify in experimental data factors that negatively affect passengers’ attitudes towards airline services. When conducting the study, well-known algorithms were used that are part of free WEKA (Waikato Environment for Knowledge Analysis) software for data analysis and machine learning by University of Waikato (New Zealand), distributed under the GNU GPL license: naïve Bayes classifier; multilayer perceptron using backpropagation algorithm; k-nearest neighbour method (KNN) with adaptive selection of parameters; decision trees – J48 is an open-source Java implementation of the C4.5 algorithm; random forest; logistic regression; adaptive boosting algorithm (AdaBoost); support vector machine – the SMO (Sequential Minimal Optimization) algorithm which is one of the possible implementations of the support vector machine algorithm. It is shown that the best accurate models reflecting passenger satisfaction with airline services are obtained using random forest algorithms (error on the test sample is of 3,9 %) and a neural network approach (error on the test sample is of 3,7 %). At the same time, these algorithms do not allow us to explicitly identify factors characteristic of air passengers who are dissatisfied with the quality of service. This gap is filled by an algorithm based on the method of structural resonance in multidimensional data (SRMD), which made it possible to identify precise logical rules in the data with high completeness. The resulting logical rules are highly interpretable patterns of passengers who either negatively or neutrally evaluate the airline’s services in general.
Authors: V. A. Duke, I. G. Malygin
Institution: Russian Academy of Sciences
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4405599616