Alif Dwi Kurniawan, Heni Vidia Sari, Triyanna Widiyaningtyas, Muhammad Anandha Fritama, Andriana Kusuma Dewi, Laili Hidayati
In recent decades, the video game industry has grown massively. This has resulted in a large amount of content being available, necessitating the grouping of game content according to user needs. This research aims to classify age categories in video games, so users can obtain information about the suitability of a game for the expected age group. The classification method used in this research is Support Vector Machine (SVM). Testing was conducted on the video game rating dataset by ERSB (Entertainment Software Rating Board) using the multiclass SVM One-vs-Rest and One-vs-One methods, and applying several schemes for comparing training data and test data (i.e., 60:40, 70:30, 80:20, and 90:10). The experimental results show the highest accuracy value based on evaluation using a confusion matrix in the multiclass SVM One-Vs-Rest method with a 90:10 data division scheme, achieving an accuracy of 91.57%, precision 92%, and recall 92%. Content descriptors such as 'blood' and 'fantasy violence' were examined to understand their impact on age category classifications. This shows that increasing training data is able to improve accuracy, with a training data and test data scheme of 90:10 having the best performance. © 2024 IEEE.
Universitas Negeri Malang, Department of Electrical Engineering and Informatics, Malang, Indonesia; Universitas Negeri Malang, Department of Culinary and Fashion Education, Malang, Indonesia