Content Based Hotel Recommender Based on Multi-Aspect Sentiment Analysis by Rating Inference from User Reviews

Introduction: In today’s era of e-business, the online purchase of goods and services has become an essential and inseparable part of human life. Recommender systems play a crucial role in helping users select products or services that are both relevant and reasonably priced. Objectives: Current recommender systems generate product or service recommendations by identifying similarly rated items and matching them with other similar users. However, these systems do not consider user preferences for specific product features and rely solely on single ratings given by other users. The objective of this study is to propose a more accurate recommender system that improves recommendation precision by analyzing multiple aspects of user preferences and increases the personal trust. Methods: The proposed recommender system enhances accuracy by analyzing user behavior and activities toward different products. It mines user opinions for multiple product aspects by utilizing physical ratings provided for each hotel and determining sentiment scores from textual reviews. Additionally, the system infers new scores from sentiment analysis. Various machine learning classification techniques, such as Naïve Bayes, maximum entropy, Support Vector Machine (SVM), Decision Tree, and Random Forest, are employed along with functional units like feature extraction and sentiment analysis to improve recommendation effectiveness. Results: The final set of recommendations is prepared by considering multiple aspects of the problem domain rather than relying on single ratings. This approach leads to a more personalized and accurate recommendation system. The proposed system has been tested, and its accuracy outperforms existing similar systems by 2% to 9%. Conclusions: By incorporating multiple aspects of user opinions and preferences through advanced machine learning techniques, the proposed recommender system achieves higher accuracy compared to traditional models. The improvements in accuracy demonstrate the system's effectiveness in providing more precise and user-centric recommendations.