Web Service Classification From Network Flow Using Undersampling Technique

Web service classification is crucial in managing many client requests for various applications. One of the critical challenges in web service classification is the high variability and diversity of web services. To address this challenge, this paper proposes a decision tree method enhanced with undersampling for web service classification. This method selects each class in the dataset with a target of 10,000 data instances per class. Any additional classes beyond this threshold will be removed from the data set. This undersampling technique helps create a more balanced representation of each class, ensuring that all classes contribute equally during the classification process. By incorporating undersampling with a class selection approach, the Random Forest model can effectively handle class imbalance and significantly improve its accuracy and efficiency in web service classification. The experiment conducted in this study yielded promising results, with the Random Forest approach achieving an accuracy of 82%. This performance outperformed other classification algorithms such as Naïve Bayes, K-Nearest Neighbors (KNN), Decision Tree, and Multilayer Perceptrons.