Social Media Sentiment Analysis for Airline Customer Satisfaction

Among the most important subfields in natural language processing, one of the fields is sentiment analysis. Sentiment analysis focuses on recognizing and estimating emotions in text. It has gained more importance in aviation in the sense that it can present insights into customer satisfaction and brand reputation as well as operational efficiency. This work investigates the topic of the sentiment analysis of the Twitter US Airline Sentiment Dataset: more than 14,000 tweets regarding the major US airlines labelled as positive, neutral, or negative. By such approaches documented in the literature, challenges consisting of the noise and subtlety of sentiment in the given data largely remain unsolved. To address these deficiencies, this research will use the machine learning techniques in the form of Naive Bayes and Random Forest models, classifying sentiments by methods like class imbalances through SMOTE, and using TF-IDF vectorization in extracting features. Results for the experiment showed that the Random Forest model outperforms Naive Bayes as the former can classify more neutral and positive sentiments correctly. It is, therefore, concluded that sentiment analysis may be useful to airlines for enhancement in customer experience and future research should emphasize on advanced deep learning models with rich data sets for further potential enhancement.