Enhancing Travel Recommendation Using Collaborative Filtering and Hybrid Data-Driven Architecture

Generally, Recommender Systems (RS) are used to address the problem of information overload in a particular domain. When it comes to travel recommender, travelers often face the issue of excessive effort required to find and select their preferred travel options from a large, unorganized pool. Thus, planning vacations can be a time-consuming and daunting task, requiring extensive research and decision-making. While traditional recommendation systems achieved notable success, they also encountered various challenges. The emergence of deep learning models, particularly in the field of computer science, presents a significant opportunity to enhance recommendation systems. In our work, we have aimed to lessen the efforts of vacation planning and enable travelers to spend more time enjoying their trips by collecting information such as destination, travel dates, budget, hotel amenities, preferred attractions, and cuisine preferences from users, using a comprehensive dataset. Moreover, we have obtained significant accuracy using three recommendation models for three different recommendation systems (Attractive places to visit, Hotels, and Restaurants)– two collaborative filtering models (Restricted Boltzmann Machine (RBM) and Matrix Factorization using Alternating Least Squares (MF-ALS)) and one machine learning based hybrid architecture, combines K-Means and KNN. We’ve separately evaluated each model’s performance using several performance metrics including Accuracy, Precision, Recall, and F1-score. Then we compared our models with some state-of-the-art research and our selected ones outperformed others in the recommendation set. In this approach, it is highlighted that the effectiveness of the integrated collaborative filtering and hybrid data-driven infrastructure is a strong recommender for travel recommendation systems.