Tourist Spot Recommendation from Images based on Age Group and Location for Dubai using Deep Transfer Learning

Tourism is an important sector which can boost the economy of a country. Tourist recommender systems are used to provide recommendations to tourists based on their preferences. The prime challenge a tourist face while visiting tourist spots in a country is in determining a tourist spot to visit next that best suits their age. This paper proposes a deep learning-based system called TSR-Net that recommends next tourist spot from images of a tourist in Dubai by predicting the age group and the current tourist spot. TSR-Net uses two pretrained MobileNet models for this purpose. One MobileNet predicts the age group from the image and the other MobileNet identifies the current tourist spot from the image. Once the age group and the current tourist spot are identified the next suitable tourist spot is recommended to the tourist. The TSR-Net is applied to datasets to evaluate its performance on age group classification (AGC) and tourist spot detection (TSD). Results proveed that the proposed model achieved the best accuracy of 82% and 100% for AGC and TSD. The proposed model was also tested on a custom dataset and showed satisfactory results. The TSR-Net classifies a person into one of 3 age groups (less than 18, 18-45, more than 45), identifies 4 tourist spots in Dubai, UAE (Burj Khalifa, Burj Al Arab, Atlantis, Ain Dubai) and recommends 20 different places. Thus, the proposed model will help tourists visit appropriate places according to their age and present location.