Abstract

The proliferation of user-generated reviews on social media and travel platforms has created unprecedented
opportunities for modelling tourist behaviour using machine learning techniques. Yet, existing studies rarely
integrate sentiment-derived features with full-spectrum rating classes, limiting their predictive capacity in
real-world tourism analytics. This study proposes a sentiment-enhanced Multilayer Perceptron Neural
Network (MLP-NN) model for classifying tourist ratings (1–5 stars) based on textual reviews of the Petronas
Twin Towers sourced from TripAdvisor. A dataset of 1,850 reviews was extracted, cleaned, and augmented
with polarity, subjectivity, and confidence scores generated through Aylien’s sentiment analysis engine.
These engineered features were used to train and validate the MLP-NN within a supervised learning
framework. Experimental results show that the proposed model achieves an accuracy of 86.1%,
outperforming Support Vector Machine and K-Nearest Neighbour benchmarks by approximately 19%. The
findings demonstrate that coupling sentiment metrics with neural network architectures substantially
enhances predictive performance, offering a scalable approach for anticipating tourist satisfaction and
informing evidence-based destination management. This study contributes a robust methodological
framework that can support tourism marketers, policymakers, and platform designers in leveraging large-scale review data for strategic decision-making and competitiveness in a data-driven tourism ecosystem.

Key words: Machine Learning; Tourists Review; Classification; Multilayer Perceptron; Artificial Neural Network