Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Mining the Skies: An Exploration of Airline Reviews using LDA

Ljungström, Joel LU (2023) DABN01 20231
Department of Economics
Department of Statistics
Abstract
Airlines face a highly competitive industry requiring large capital investment at thin profit margins. In addition, the carriers are re-emerging from the COVID-19 pandemic; a time period which brought the industry to a standstill and revenue streams to an unforeseen low. Airlines are now facing their next challenge: regaining their market shares to pre-pandemic levels, and beyond. As such, reputation and recommendations are important factors for the companies to gain new customers, and online reviews presents a unique, low-cost opportunity to learn from indirect feedback and predict what areas of their products and services leads to positive recommendations. Through Latent Dirichlet Allocation (LDA), this study extracted 18 unique latent... (More)
Airlines face a highly competitive industry requiring large capital investment at thin profit margins. In addition, the carriers are re-emerging from the COVID-19 pandemic; a time period which brought the industry to a standstill and revenue streams to an unforeseen low. Airlines are now facing their next challenge: regaining their market shares to pre-pandemic levels, and beyond. As such, reputation and recommendations are important factors for the companies to gain new customers, and online reviews presents a unique, low-cost opportunity to learn from indirect feedback and predict what areas of their products and services leads to positive recommendations. Through Latent Dirichlet Allocation (LDA), this study extracted 18 unique latent topics from 128,631 samples to identify key areas often written about in reviews. Additionally, the probabilities of these topics to occur in reviews were used to predict the outcome of recommendation and overall ratings with the use of classification trees and variations of logistic regression. The top performing model had an accuracy of 85.77% in predicting recommendation, and multiple areas were identified as opportunities for airlines to make managerial decisions on to improve their reputation online. Key dimensions relating to a positive recommendation found were Good Service, Efficiency and Cabin Crew, whereas dimensions relating to a negative recommendation were identified as Bad Customer Service, Travel Delays, and Charges. (Less)
Please use this url to cite or link to this publication:
author
Ljungström, Joel LU
supervisor
organization
course
DABN01 20231
year
type
H1 - Master's Degree (One Year)
subject
keywords
LDA, Logistic Regression, Classification Tree, Airline Reviews, Web Crawler
language
English
id
9119169
date added to LUP
2023-11-21 12:54:40
date last changed
2023-11-21 12:54:40
@misc{9119169,
  abstract     = {{Airlines face a highly competitive industry requiring large capital investment at thin profit margins. In addition, the carriers are re-emerging from the COVID-19 pandemic; a time period which brought the industry to a standstill and revenue streams to an unforeseen low. Airlines are now facing their next challenge: regaining their market shares to pre-pandemic levels, and beyond. As such, reputation and recommendations are important factors for the companies to gain new customers, and online reviews presents a unique, low-cost opportunity to learn from indirect feedback and predict what areas of their products and services leads to positive recommendations. Through Latent Dirichlet Allocation (LDA), this study extracted 18 unique latent topics from 128,631 samples to identify key areas often written about in reviews. Additionally, the probabilities of these topics to occur in reviews were used to predict the outcome of recommendation and overall ratings with the use of classification trees and variations of logistic regression. The top performing model had an accuracy of 85.77% in predicting recommendation, and multiple areas were identified as opportunities for airlines to make managerial decisions on to improve their reputation online. Key dimensions relating to a positive recommendation found were Good Service, Efficiency and Cabin Crew, whereas dimensions relating to a negative recommendation were identified as Bad Customer Service, Travel Delays, and Charges.}},
  author       = {{Ljungström, Joel}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Mining the Skies: An Exploration of Airline Reviews using LDA}},
  year         = {{2023}},
}