Adaptive Forensic Analysis of Network Traffic Using Advanced Machine Learning Gradient Boosting

Authors

  • Mukhlis Hanafi
  • Alam Rahmatulloh

DOI:

https://doi.org/10.32764/23m7e695

Keywords:

Computation Framework, Harmful Pattern, Descriptive attribute, Predictive stability, Distribution variation, Operational monitoring system

Abstract

This study presents a dynamic computational framework designed to strengthen the identification of harmful patterns within complex communication environments by employing an adaptive decision-tree ensemble capable of adjusting to changes in data characteristics. Using a widely recognized benchmark from 2017, the investigation demonstrates how meticulous preparation of traffic observations, refinement of descriptive attributes, and balanced scaling contribute to improved predictive stability. The adaptive model constructs trees sequentially while modifying its parameters in response to distributional variations, enabling sustained performance under evolving conditions. Quantitative evaluation shows that the method reaches an accuracy of 0.6651, precision of 0.6012, recall of 0.9974, F1-score of 0.7502, and an area-based measure of 0.9866, consistently surpassing non-adaptive counterparts across all indicators. Robustness tests under induced behavioral shifts further confirm its resilience, while validation through internal and external procedures highlights its reliability across different traffic contexts. The analysis reveals strong responsiveness, high interpretability, and stable predictive capability, underscoring the importance of adaptive learning in environments defined by irregular behavior and unpredictable activity patterns. The findings also offer practical insight for operational monitoring systems that require continuous adjustment to maintain effective identification of undesirable events. Future work may enhance this framework by integrating deeper contextual descriptors, optimizing real-time deployment, and addressing computational demands of large-scale adaptive processes.

Downloads

Download data is not yet available.

Downloads

Published

2026-11-15

Issue

Section

Articles

How to Cite

Adaptive Forensic Analysis of Network Traffic Using Advanced Machine Learning Gradient Boosting. (2026). NEWTON: Networking and Information Technology, 6(2), 69-79. https://doi.org/10.32764/23m7e695