Classification of User Reviews on Online Travel Booking Applications in Bangladesh Using Multinomial Naive Bayes

The exponential development of technology in Bangladesh has changed the face of the tourism sector, and even with the inclusion of application-based online travel ticket booking systems like Shohoz, GoZayaan and BDTickets.These apps are popular among Bangladeshi travelers.In this study, we examine an architecture where the Multinomial Nave Bayes algorithm is used to classify user reviews of these travel apps into two categories, which are "Satisfied" and "Unhappy".The dataset is composed of 1339 reviews, which were gathered from the Google Play Store.The results were considered, considering the data was split into three scenarios (70:30, 80:20, 90:10 and validated using a confusion matrix and KFold Cross Validation.Among all models, 81.34% accuracy was obtained for the 9:1 split ratio with the model precision of 81.47%, recall of 81.43%, and F1-Score of 81.34%.A TF-IDF analysis showed that terms like "good," "nice," and "excellent" were the most prevalent in the "Satisfied" class, whereas terms such as "price," "can't," and "app" were more likely to be found in the "Unhappy" class.The results indicate that the Multinomial Nave Bayes approach is efficient for the classification of user reviews of online travel booking applications in Bangladesh, and the performance increases as the data table size increases.