Framework on Image-Based Implicit Tourism Recommendation Using Transfer Learning Feature Extraction

Tourism is one of the leading sectors that has a huge impact on micro and creative industries. People typically have difficulty making decisions when there is either too little or too much information regarding tourist attractions. This is when a recommendation system comes into play. In this paper, we proposed and designed an image-based tourism destination recommendation system framework. The aim of the proposed framework is to enable the implicit recommendation which takes source information from the selected, clicked, or liked photo images. To deliver the recommendation, we use transfer learning with VGG16 as the feature extraction architecture, Principal Component Analysis (PCA) to reduce the number of features, and the cosine similarity technique to measure the matching-ness between the query data and the tourism destination objects. In our initial framework, we used 57 tourist destinations in Yogyakarta, Indonesia as the recommendation objects. The recommendation resulting from the framework has been evaluated by users, and we got 96% of the respondents who agree that the framework gives suitable recommendation from the query image given. Thus, the proposed framework can help users find similar tourist attractions based on their preferred past photo.