A multi-criteria recommendation system for personalised tourism experiences with user query analysis

Abstract

Recommendation systems play a crucial role in assisting users with decision-making by suggesting relevant items. Multi-criteria recommender systems (MCRS) enhance this process by incorporating user preferences for various aspects, leading to more personalised and effective recommendations. However, MCRS faces challenges such as high computational complexity and limited consideration of user context, including user preferences for relaxation, which may differ between solo trips and trips with friends. This paper addresses these limitations by proposing a novel MCRS approach for tourism recommendation systems. Our proposed system combines matrix factorisation with a deep residual network (ResNetMF), demonstrating substantial performance improvements in terms of RMSE, MAE and lower training time compared to a wide range of baselines. Additionally, a user query analysis component allows users to express their dynamic preferences through queries, catering to the context-specific nature of travel decisions. The evaluation demonstrates that our ResNetMF model outperforms baseline and deep learning methods in most of the tested evaluation metrics while having the lowest training time. This work contributes to the field of tourism recommendation systems by proposing a user-centred approach that addresses both accuracy and user interaction for effective travel recommendations.