| Popis: |
Social media, such as Facebook and Twitter, has become an indispensable part of the Internet. These online social platforms facilitate the information exchange between people. At the same time, some users create multiple accounts on the same social media, pretend to be others, and publish articles or comments through different accounts to influence online public opinion. We refer to multiple accounts created by the same user named as sockpuppet accounts. In the past, people's research on sockpuppet detection was mostly based on verbal features or non-verbal features. These kinds of methods have their unique advantages and drawbacks, most of them ignore the connection between the two methods and study separately. For this purpose, we propose an adaptive multi-source feature fusion method, which not only considers the importance of text in identifying the sockpuppet, but also considers the characteristics of users’ online behavior which is not easy to disguise. The experimental results on real datasets show that our method has a better detection rate than compared with competitive algorithms. |
