Fuzzy ontology as a basis for recommendation Systems for Traveler’s preference

As an increasingly larger number of users partake in Facebook, the images shared and posted by user in Facebook provide a richer background to build a more accurate recommendation system. In this paper, we propose an Intelligent Recommendation System for Travelers’ Preferences (IRSTP) based on the conceptualization of the travel user’s destination based on their activities of sharing visual data and locations through social network. In this system, we performed a Deep Neural Network Architecture transferred on Places356 database to extract important concepts found in the shared visual images. Then, we carried out an inference system based on crisp (DOVUIS) and fuzzy ontology (DFOVUIS) to take into account some ambiguity in Decision making. Both proposed ontologies are tested and evaluated on collected Sudanese Database. The User interest architecture based on Convolutional Neural Network is used for tuning search in the IRSTP recommendation system. The Fuzzy ontology performs well than the crisp inference system in order to personalize the recommendation based on the profile outputted from this ontology. We achieve an accuracy of 94.22\% on the Sudanese database to classify the topic of user interest among 9 classes Food (98.7%), Nature (98.9%), History (80.4%), DIY and Craft (98.0%), Celebrities (87.3%), Architectures (98.6%), Events (95.4%), Holydays (93.9%) and Art (87.0%). These topics are considered from the Facebook generic topics.