N-Gram Keyword Retrieval on Association Rule Mining for Predicting Teenager Deviant Behavior from School Regulation

摘要

Nowadays teenagers are affected by various advances in technologies. Hence they need to have the ability to handle the challenge and maintain good behavior especially during their studies. Parents and education institutions have been focusing on helping students to excel in studies and organizational skills. However, deviant behaviors could occur during adolescence and affect students learning performance. Association rule mining is implemented in this research to predict teenagers deviant behaviors at the earliest based on previous violation cases of junior high school regulations. Furthermore, we can input text of problem in Indonesian language which contains at least one type of deviant behavior to predict future trouble possibilities to prevent them from happening beforehand. The system developed consists of two main phases which are the keyword extraction phase and association rule mining phase. The keyword extraction phase includes filtration, stemming, and keywords retrieval with n-gram and tf-idf. The association rule mining phase includes keywords tree construction, data transformation, association rules analysis, and application of discovered knowledge. Association rules and the predicted deviant behavior are the output of the system. Our experiment results in 80% precision score and 72% recall score.