A Study on Photo and Tag-based Recommendation Methods for the Homestay Industry

• Main Authors: Wu, FuHsien, 吳馥先
• Other Authors: Wu, JiTsung
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/17220284390740039198

碩士 === 輔仁大學 === 資訊管理學系 === 100 === Most people choose homestay based on the photo on the website which will also affect the housing rate. Su (2010) explored the user preferences of tag browsing and the use of recommended methods. Zhang (2011) mentioned that the social tagging is considered to be able to render better homestay diversity. Based on the results of Su(2010) and Zhang(2011), this study investigates the effectiveness of User-homestay Matrix、TF-IDF and Cosine Similarity for tag and photo based homestay recommendation. This study conducted an online experiment. First, let user chose the picture for five rounds. The system will record user profile with equal weight, so that each label weights of user click is the same. Second, system uses the three recommended methods to calculate. There are user-homestay Matrix, TF-IDF and Cosine similarity. Each method calculates the two highest scores of Homestay and filters out the duplicate homestay for the user to score. The experimental results show the precision of Cosine similarity is the highest and recommended method of TF-IDF is only a difference of 4%. The user-homestay Matrix method recommended significantly worse than other methods. After the end of the experiment, the survey on the subjects’ occupancy found that the accuracy of the system would affect the subjects’ occupancy. The accuracy of the recommendation system has a considerable influence on the choice of homestay.