Tourism has become one of the world’s largest service industries. Due to the rapid development of social media, more people like self-guided tours than package itineraries planned by travel agencies. Therefore, how to develop itinerary recommendation systems that can provide practical tour suggestions for tourists has become an important research topic. This study proposes a novel tour recommendation system that considers the implicit and dynamic information of Point-of-Interest (POI). Our approach is based on users’ photo information uploaded to social media in various tourist attractions. For each check-in record, we will find the POI closest to the user’s check-in Global Positioning System (GPS) location and consider the POI as the one they want to visit. Instead of using explicit information such as categories to represent POIs, this research uses the implicit feature extracted from the textual descriptions of POIs. Textual description for a POI contains rich and potential information describing the POI’s type, facilities, or activities, which makes it more suitable to represent a POI. In addition, this study considers visiting sequences when evaluating user similarity during clustering so that tourists in each sub-group hold higher behavior similarity. Next, the Non-negative Matrix Factorization (NMF) dynamically derives the staying time for different users, time slots, and POIs. Finally, a personalized itinerary algorithm is developed that considers user preference and dynamic staying time. The system will recommend the itinerary with the highest score and the longest remaining time. A set of experiments indicates that the proposed recommendation system outperforms state-of-the-art next POI recommendation methods regarding four commonly used evaluation metrics.
A multiple objective simulation optimization algorithm named Multiple Objective Probabilistic Branch and Bound with Single Observation (MOPBnB(so)) is presented for approximating the Pareto optimal set and the associated efficient frontier for stochastic multi-objective optimization problems. MOPBnB(so) evaluates a noisy function exactly once at any solution and uses neighboring solutions to estimate the objective functions, in contrast to a variant that uses multiple replications at a solution to estimate the objective functions. A finite-time performance analysis for deterministic multi-objective problems provides a bound on the probability that MOPBnB(so) captures the Pareto optimal set. Asymptotic convergence of MOPBnB(so) on stochastic problems is derived, in that the algorithm captures the Pareto optimal set and the estimations converge to the true objective function values. Numerical results reveal that the variant with multiple replications is extremely intensive in terms of computational resources compared to MOPBnB(so). In addition, numerical results show that MOPBnB(so) outperforms a genetic algorithm NSGA-II on test problems.