A simulated annealing-based recommender system for solving the tourist trip design problem

Tourism recommender system (TRS) experienced high growth over the last decades given its significant role in tourist satisfaction. Tourists are usually worried about planning their trips since it requires a considerable effort and it is a time consuming task. The main purpose of TRS is to select points of interest (POIs) that match users preferences and to suggest personalized daily sightseeing Designing such a suitable traveling tour is modeled as a Tourist Trip Design Problem (TTDP). Numerous variants of the problem are presented in the literature such as the Team Orienteering Problem (TOP) evoked in this paper. The aim is to maximize the total collected score of visited POIs subject to distance and time constraints. For solving the TTDP, we propose a metaheuristic based on two steps (i) grouping the POIs into clusters based on the start and end locations and (ii) solving the routing problem for each cluster to generate the optimal routes. The proposed approach is a combination between k-means and simulated annealing algorithms, namely, KSA. Besides, we develop StayPlan which is a TRS that integrates KSA algorithm to tackle the TTDP. StayPlan incorporates three main modules which are, user preferences, data retrieving & scoring and TTDP solving & optimization. Computational experiments compare the performance of our proposed algorithm with four state-of-the-art methods. The results show the efficiency and effectiveness of KSA algorithm in solving the benchmark suite from the literature. KSA algorithm reaches the best-known solutions for 100% of tested instances. Additionally, our method outperforms all the considered algorithms in terms of CPU time, it is quite fast with an average run time of 31.49 s. Ultimately, an illustrative example of the adapted scenario is applied to Tunis city as a real case.