Forecasting airline passengers’ satisfaction based on sentiments and ratings: An application of VADER and machine learning techniques

Abstract

To the best of the authors' knowledge, research predicting airline passengers' satisfaction based on their sentiments and ratings is seldom sighted. Additionally, the literature reveals that most studies have primarily concentrated on specific airlines or routes, neglecting to conduct a comparative analysis of satisfaction levels across numerous airlines and routes. Hence, this research aims to predict passengers' satisfaction by combining the sentiment of their reviews and ratings on various parameters like food, entertainment, seat comfort, ground service, and value for money. Using the "Skytrax Airline Reviews" dataset, which contains data about 81 airlines and 64440 reviews, our research analyzes and predicts airline passengers' satisfaction based on sentiments and ratings using nine popular machine learning techniques. The study found that the LightGBM obtains an accuracy of 97 percent in predicting customer satisfaction. The results further reveal that 'value for money' and 'ground service' are crucial factors in determining the passengers' satisfaction, whereas 'entertainment' had no significant impact. Our study thus provides a valuable tool for predicting airline industry customer satisfaction and gives insight into the factors contributing to passenger satisfaction. These findings can further help airlines better understand their passengers' needs and improve their services accordingly.