<![CDATA[Traffic anomaly detection is a critical element in maintaining computer network security against threats such as DoS, Brute Force, and SQL Injection. However, the effectiveness of intrusion detection systems heavily relies on the selection of optimal classification algorithms and appropriate data preprocessing techniques. This study aims to compare the performance of Naive Bayes, C4.5, and Random Forest data mining algorithms, and analyze the impact of feature selection using Information Gain on detection accuracy. The experimental method employs the CICIDS2017 WebAttack subset dataset consisting of 170,366 records, with evaluation based on accuracy, True Positive Rate (TPR), False Positive Rate (FPR), and ROC Area metrics. Results indicate that the Random Forest algorithm achieved the best performance with 100% accuracy in use training set mode without feature selection. A significant finding reveals that the application of Information Gain feature selection drastically decreased classification performance, particularly in the Naive Bayes algorithm which experienced an accuracy drop to 23%. It is concluded that for the CICIDS2017 dataset characteristics, using all initial features is recommended over feature reduction, with Random Forest being the most effective algorithm for traffic anomaly detection system implementation]]>
Copyrights © 2026
