상세정보
학술대회
Social Link Prediction and Feature Analysis in Mobile Game
- 발행일
- 201810
- 출처
- International Conference on Information and Communication Technology Convergence (ICTC) 2018, pp.906-909
- 협약과제
- 18CS1300, 지능형 라이브 서비스를 위한 게임 운영 시나리오 최적화 플랫폼 기술 개발, 양성일
- 초록
- Predicting, monitoring, and analyzing player's behavior is essential for free-to-play mobile game business. Especially, game company have made their game with social features in order to prevent users from leaving. In this paper, we predict social relationship from game data and analyze factors affected to prediction accuracy by features, prediction period, machine learning algorithm, and social properties. The experiment is performed on commercial mobile game. Random forest shows the best performance. Psychological factors have small effect on link prediction and last 7 days within training period affects the most of future relationship.
- KSP 제안 키워드
- Best performance, Feature Analysis, Free-To-Play, Link prediction, Machine Learning Algorithms, Mobile Game, Prediction accuracy, Prediction period, Psychological factors, Random forest, Social Properties