Implementasi Locally Adaptive K-Nearest Neighbor Algorithm based on Discrimination Class (DC-LAKNN) pada Kasus Deteksi Fake Account Instagram

Yosefani Kurniawan (Program Studi Teknik Informatika, Universitas Kristen Petra Surabaya)
Lily Puspa Dewi (Program Studi Teknik Informatika, Universitas Kristen Petra Surabaya)
Silvia Rostianingsih (Program Studi Teknik Informatika, Universitas Kristen Petra Surabaya)

Jurnal Infra
Vol 10, No 1 (2022)

Instagram is one of the social media that has many users. Because of the ease of creating an account, many people create a fake account for stalking, spam attempts, fraud, photo or password theft, and even attacks another account with virus. Therefore, users need to be wary of unknown followers. Detecting account, which is real or fake can help users to be careful accepting some unknown follower. In addition, users can report to Instagram so that account can be deactivated. In this thesis, a website-based application is designed that can detect the possibility of an Instagram account being a real or fake account. The detection is carried out using the Locally Adaptive K-Nearest Neighbor algorithm classification method based on Discrimination Class (DC-LAKNN) which is an adaptive algorithm from the K-Nearest Neighbor algorithm. This algorithm pay attention at discrimination class as the basis for classification. The attributes used in the classification are user follower count, following count, biography length, media count, username digit count, username length, user has profile picture, user is private. The end result is that the Locally Adaptive K-Nearest Neighbor algorithm based on Discrimination Class (DC-LAKNN) can be used to classify Instagram accounts with an accuracy of 96.23%.