Comparative analysis of k-NN and DT model in sentiment classification of Labuan bajo-wonderful Indonesia content reviews

Abstract

This research investigates the efficacy of sentiment classification models, specifically k-NN and DT algorithms, in the context of destination branding, with a focus on Labuan Bajo tourism. Utilizing the CRISP-DM (Cross-Industry Standard Process for Data Mining) framework, the study systematically navigates through all six stages, including business understanding, data understanding, data preparation, modeling, evaluation, and deployment, to analyze textual reviews and gauge public sentiments towards Labuan Bajo. The findings reveal that both k-NN and DT models exhibit high accuracy and precision, with k-NN achieving an average accuracy of 97.79% and DT 97.52%. While k-NN demonstrates commendable performance in recall, DT exhibits superior discriminative power, particularly when integrated with SMOTE, as evidenced by higher AUC values. The research underscores the importance of leveraging advanced machine learning techniques for sentiment analysis to inform destination branding strategies effectively. These insights provide valuable guidance for stakeholders in enhancing the branding and promotion of Labuan Bajo as a premier tourist destination, ultimately contributing to its sustainable development and global recognition