@techreport{oai:ipsj.ixsq.nii.ac.jp:00185063,
 author = {Bin, Du and Wenbo, Su and Mizuho, Iwaihara and Bin, Du and Wenbo, Su and Mizuho, Iwaihara},
 issue = {14},
 month = {Dec},
 note = {Users from the entire world interact with others, share feelings and do various activities on SNSs. Such activities of users potentially have strong relationships with their interests and preferences. In this paper, we investigate relationship between public posting activities, online behaviors of users and their belonging interest groups, to identify representative user types, and compare distribution of these user types in each interest group. Public posts of users in several Facebook interest groups have been collected, and linguistic features such as the length of each post have been extracted. The polarity and the subjectivity of posts have been obtained through sentiment analysis. Combined with activity features of these users, post-level clustering and user-level clustering are carried out by several clustering algorithms, and six user types representing these clusters are identified. We observe that distribution of these user types is distinctively different between interest groups, and this fact is useful for predicting users' interests., Users from the entire world interact with others, share feelings and do various activities on SNSs. Such activities of users potentially have strong relationships with their interests and preferences. In this paper, we investigate relationship between public posting activities, online behaviors of users and their belonging interest groups, to identify representative user types, and compare distribution of these user types in each interest group. Public posts of users in several Facebook interest groups have been collected, and linguistic features such as the length of each post have been extracted. The polarity and the subjectivity of posts have been obtained through sentiment analysis. Combined with activity features of these users, post-level clustering and user-level clustering are carried out by several clustering algorithms, and six user types representing these clusters are identified. We observe that distribution of these user types is distinctively different between interest groups, and this fact is useful for predicting users' interests.},
 title = {Relationship between SNS users' posting activities and their interests},
 year = {2017}
}