KLASIFIKASI KUALITAS LAYANAN JARINGAN NIRKABEL KAMPUS BERDASARKAN PARAMETER QUALITY OF SERVICE MENGGUNAKAN ALGORITMA RANDOM FOREST DAN GRADIENT BOOSTING

Abstract

Campus wireless network (Wi-Fi) infrastructures play an indispensable role in empowering digital learning, academic information system interactions, and institutional research. Nevertheless, severe client concurrency during peak lecture hours frequently causes Quality of Service (QoS) degradation manifested in excessive latency, jitter spikes, and packet loss. This paper develops automated wireless service quality classification grounded upon TIPHON and ITU-T G.1010 standards through a comparative exploration of modern ensemble learning: Random Forest (bagging) and Gradient Boosting Decision Trees (boosting). The experimental dataset encompasses 1,200 network telemetry records monitoring Packet Loss, Latency, Jitter, Throughput, RSSI, and SNR categorized across four quality classes (Poor, Moderate, Good, and Excellent). Employing stratified 10-fold cross-validation, Random Forest achieved optimal classification accuracy of 100.00% with a 100.00% F1-score. Gradient Boosting yielded 99.67% accuracy and 99.67% F1-score. Feature importance analysis confirms that Latency (33.41% on RF and 55.98% on GBDT) and Throughput (22.97% on RF and 40.29% on GBDT) are the paramount determinants for wireless link optimization.