Unveiling Delta Airline Experiences through Topic Modeling and Traveler Persona Classification

Understanding customer feedback is crucial for enhancing service quality in the competitive airline industry. This study analyzes Delta Airline's customer reviews using advanced natural language processing techniques, focusing on topic modeling and traveler persona classification. Initially, the study was performed an elbow plot analysis to determine the optimal number of clusters, identifying three distinct clusters. Clustering algorithm K-means and Latent Dirichlet Allocation were then applied to uncover the main topics in the reviews. Next, this research classifies reviewers into different traveler personas using NLP models such as BERT, Word2Vec, One-Hot Encoding, DistilBERT and RoBERTa, and models which uses deep learning architectures such as LSTM, BiLSTM and GRU identifying categories like Family leisure, solo leisure, Couple leisure and business travelers. The insights from this analysis enable Delta Airline to tailor their services to specific customer segments. Classification results indicate that BiLSTM and GRU models performed the best, with BiLSTM achieving an accuracy of 92% with Roberta.