[1J2-J-6] Web mining 1

[1J2-J-6-03] User Identification Across Social Media Based on Friends' Locations

〇Kazufumi Kojima¹, Masahiro Tani¹ (1. NEC Corporation)

Keywords:User Identification, Social Media

This paper describes user identification method across social media using friends' location information. One of the conventional methods is based on the similarity of two display names. The conventional method has a problem of a decrease in the identification accuracy if a user registers different display name across social media. Our proposed method aims to address this problem utilizing friends' location information. We convert location information extracted from social media to that of administrative unit such as country and state, and calculate the weighted occurrence frequency of a location pair based on distance between the locations. This basis is that a friend list rarely includes the account pair whose locations are far distance. Finally, we identify users using the account similarity between weighted occurrence frequencies on each administrative unit. The evaluation experiment shows that the proposed method improves the performance rather than the conventional method(accuracy:82.7%→90.9%).