HYBRID RECOMMENDER SYSTEM FOR TOURIST DESTINATION IN INDONESIA USING USER INSTAGRAM DATA

• Main Author: Muthmainnah
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/43812
• Institution: Institut Teknologi Bandung

Tourism is a leading sector in Indonesia, but unfortunately the wide range of selection for various tourist destinations in Indonesia creates overwhelming information for tourists and makes it hard for them to choose the best destination that fits their preferences. On the other side, tourists share their travel experiences through social media Instagram. These tourists’ Instagram data later being used for developing recommender system using collaborative and content-based method. Collaborative recommendations are created using Instagram users’ travel history, while content-based recommendations are created by matching user’s profile with destination’s profiles which are built based on what hashtags are tagged on Instagram posts found on each destination. Experiments are done for each recommendation method to conclude best model configurations. Best collaborative model configuration uses binary value to represent users’ check-in information and uses jaccard similarity metric to compute similarity between destination items. Best content-based model configuration uses binary value to represent hashtag importance on each destination and uses cosine similarity metric to compute similarity between user profile and item profile. Both recommendation models have its own advantages and disadvantages, thus hybrid method is used to collaborate those models. Based on experiment, best hybrid recommendations are produced by combining collaborative score and contentbased score with ratio 9:1. The hybrid recommendation model has been implemented on simple web application and has been tested by some users. Feedbacks given by those users imply that system’s speed is good and the accuracy of the recommendations are quite satisfying.