Predictive Classification Model dalam Tahapan Framework NIJ untuk Otomatisasi Investigasi Digital Forensik (Studi Kasus: Cyberbullying)

Abstract

This study aims to apply the National Institute of Justice  framework in the digital forensic process for conversations retrieved from LINE and Telegram applications, as well as to explore the utilization of a Predictive Classification Model  for automated text-based comment classification in cyberbullying cases. Cyberbullying is a growing form of digital crime, particularly on private and encrypted instant messaging platforms that are difficult to monitor. The research employs two machine learning algorithms within the PCM framework Complement Naive Bayes and Random Forest to detect potentially abusive comments. The forensic process follows several stages: Preparation, Evidence Assessment, Evidence Acquisition, Evidence Examination, and Documenting and Reporting, with a secure and forensically sound data extraction approach from both applications. Due to data limitations from LINE and Telegram, the classification analysis is conducted using an Instagram comment dataset that reflects the cyberbullying context. Evaluation results show that the Complement Naive Bayes model outperforms Random Forest, achieving an accuracy of 86% with balanced F1-scores, while Random Forest achieves 75% accuracy. These findings support the use of PCM as an effective aid for automatically identifying high-risk content on social media. The integration of digital forensics and artificial intelligence has significant potential to enhance the effectiveness of cyberbullying investigations.

Keywords: Cyberbullying,  Predictive Classification Model,  Complement Naive Bayes,  Random Forest,  LINE,  Telegram,  Digital Forensics, National Institute of Justice