Sentiment Analysis of Hotel Reviews Using Machine Learning: A Study on Model Performance Under Noisy Conditions

This study presents a robust, end-to-end machine learning framework for sentiment analysis of hotel reviews, designed to excel in the presence of real-world data imperfections that challenge conventional models. By integrating advanced natural language processing (NLP) techniques— noise removal, text normalization, tokenization, and lemmatization— with term frequency-inverse document frequency (TF-IDF) feature engineering, the frame work evaluates six supervised learning models: Logistic Regression, Random Forest, Support Vector Machine (SVM), K – Nearest Neighbors (KNN), Naïve Bayes, and Decision Tree. Extensive testing under varying levels of label noise demonstrates that SVM delivers superior stability and minimal accuracy degradation, making it ideal for high reliability applications. The system is deployed through a user-friendly PyQt5-based desktop application, enabling real-time sentiment predictions with confidence scores to support hotel management decision-making. This research sets a new benchmark in hospitality feedback analysis, bridging experimental AI capabilities with operational needs.