Analyzing US Airline Customer Sentiment on Twitter using Multinomial Logistic Regression and Feature Reduction

Social media has exerted a substantial impact and ongoing influence on how businesses interact with customers. Within this context, airlines have come to recognize the importance of Twitter as a pivotal avenue for connecting with customers, addressing complaints, and swiftly resolving inquiries. Employing sentiment analysis techniques, airlines can easily identify patterns, improve areas of weakness, and promptly address customer concerns. This study aims to investigate methods for improving the precision of automatic sentiment analysis of airline customers’ feedback. The proposed approach involves utilizing Term Frequency (TF), Term Frequency-Inverse Document Frequency (TF-IDF), feature selection and machine learning techniques. Based on experimental findings using the Twitter-airline sentiment database, the implementation of multinomial logistic regression based on refined TF and TF-IDF matrices has exhibited an impressive accuracy rate (81.69%).