Developing a group urban tourism recommendation system based on the modified k-means algorithm and fuzzy best-worst method

Abstract

Abstract Group recommendation is among the major concerns in urban tour guiding systems. The main challenge is the uncertainty of users’ opinions in conjunction with their preferences, which ultimately leads to the recommendation of unsuitable locations. As the number of unsuitable points of interest (POIs) for each person (tourist) increases, the efficiency of the tour guiding system faces a major decline. This paper seeks to model the uncertainty of urban tourists’ opinions regarding POIs by introducing a two-stage approach called ‘first-clustering, second recommending. The main contributions are the clustering of users based on their attributes via a modified k-means algorithm and the management of opinions using the fuzzy best-worst method (F-BWM). The proposed method is programmed for mobile applications under the name ‘G-tourism’. 485 different users registered in the mobile application and completed all the application wizard pages and 12 tours have been recognized. For each group, the POIs have been weighted, ranked, and recommended according to their members' pairwise comparisons based on BMW and F-BWM. The obtained results have been evaluated based on precision, recall, F-score measures, and user satisfaction. The accuracy assessment of running F-BWM at the second step indicates the higher accuracy of the system and user satisfaction rather than BWM.