App Recommendation Based on Matrix Factorization in Community Networks

• Main Authors: Ching-Chen Su, 蘇景宸
• Other Authors: Sung-Shun Weng
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/7ndw52

碩士 === 國立臺北科技大學 === 資訊與財金管理系碩士班 === 103 === The issue of Social Networking and Recommendation has been continuously discussed in recent years. With the prevalence of Facebook、Twitter、Google Plus and other social networks, it has been shown that the collaborative filtering approach can improve the recommendation accuracy and relieve the sparsity problem by using the information obtained through social networks. This study proposes a recommendation method applying to the community networks. The first step is to collect the rating of users and community friends for the APP items. Second, we find common preferences (similarity) among users through the collection of information, and define the similar friends with higher score as data sources. Third, by using alternating least squares, the prediction score is recommended to the target users. The experimental results show that the outcome recommended by using the approaches of similar friends and matrix factorization is accurate. Moreover, the questionnaire results show that the recommended ones are accordant with user requirements, and users are interested in the recommended outcomes.