Mas Nurul Hamidah, Triyanna Widiyaningtyas, Wahyu Sakti Gunawan Irianto, Wahyu Caesarendra
Travel recommendation algorithms on e-commerce websites are important for helping people choose the right destinations based on their preferences. However, collaborative filtering methods generally struggle with sparse data, scalability issues, and a lack of contextual awareness, which makes recommendations less accurate. This work presents the Location-Aware Recommendation System (LARES), a methodology that combines rank-based and geographical similarity through a weighted linear combination to address the sparsity problem. The spatial component includes the normalized geographical distance between users, controlled by an adjustable coefficient that balances its effect on rating similarity. The choice of public datasets from TripAdvisor and Yelp resulted in better LARES performance compared to traditional similarity models such as the Pearson Correlation Coefficient (PCC) and the EDJM method, which combines the Jeffries–Matusita distance and Jaccard Mean Squared Distance (JMSD), especially in addressing data sparsity issues. Although the Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) values for Yelp tended to be lower than those for TripAdvisor, this indicates that geographical similarity still has a strong influence on improving the accuracy and relevance of tourism recommendation systems, while introducing only limited additional computational cost. Copyright © by the authors
Department of Electrical Engineering and Informatics, Universitas Negeri, Malang, Indonesia; Universitas Bhayangkara, Surabaya, Indonesia; Department of Mechanical and Mechatronics Engineering, Curtin University Malaysia, Sarawak, Malaysia
