관광자 의사결정 지원을 위한 스마트관광 추천 시스템 연구

Abstract

The proliferation of Internet applications and digital technologies has resulted in big data in a variety of formats, including structured and unstructured data, and has created an environment for analyzing, predicting, and managing potential tourist behavior. As a result, smart tourism applications such as tourism recommender systems have advanced significantly in recent years. However, unlike the industry's development, theoretical and methodological approaches to the tourism recommender system in academia are still lacking. As a result, 'DSESTR,' a new and efficient smart tourism recommender system to aid tourists' decision-making, was proposed in this study. First, the Doc2Vec technique was used to learn the feature vectors of users and items by modeling the sequential interaction between users and items. To model the nonlinearity of consumer-item interaction, the learned vectors are fed into a multi-layer perceptron (MLP), one of the deep learning techniques. Finally, the nonlinearity learned in MLP was combined with the linearity learned in matrix factorization, which is widely used in recommender systems, to predict the user's evaluation of the item more accurately. Indeed, using the TripAdvior and MovieLens data sets, the results showed that DSESTR significantly improved prediction accuracy when compared to other recommendation algorithms. Based on the findings, the applicability of DSESTR in the tourism industry was discussed, and theoretical and practical implications were proposed.