Threshold-based Naïve Bayes classifier

• DOI: 10.1007/s11634-023-00536-8
• Publication Date: 2023-03-14T11:04:34Z
• Authors: Maurizio Romano; Giulia Contu; Francesco Mola; Claudio Conversano
• Keywords: Sentiment Analysis

Abstract The Threshold-based Naïve Bayes (Tb-NB) classifier is introduced as a (simple) improved version of the original classifier. Tb-NB extracts sentiment from Natural Language text corpus and allows user not only to predict how much sentence positive (negative) but also quantify with numeric value. It based on estimation single threshold value that concurs define decision rule classifies into opinion its content. One main advantage deriving possibility utilize results input post-hoc analysis aimed at observing quality associated different dimensions product or service or, in mirrored fashion, customer satisfaction evolve time change respect locations. effectiveness evaluated analyzing data concerning tourism industry and, specifically, hotel guests’ reviews all hotels located Sardinian region available Booking.com. Moreover, compared other popular classifiers used terms model accuracy, resistance noise computational efficiency.