How can dimensional feature reduction techniques, such as feature selection and feature extraction, be used to improve the accuracy and efficiency of botnet activity detection?
Based on our research, the effectiveness of a machine learning model depends on many factors, one of which is the features involved in the modeling. Our research also proves that dimensional reduction affects the accuracy and efficiency of the detection model. Even so, each machine learning algorithm has other factors that affect the model's effectiveness, so different feature selection and feature extraction approaches are needed.
If we look at the experimental results table, KNN and DT have different impacts even though they use the same dimensional reduction approach. In the DT, models without dimensional reduction tend to perform better than models that go through feature extraction with the concept of dimensional reduction. In contrast, the model with the KNN algorithm produces better performance when passing feature extraction with dimensional reduction. The low performance of the decision tree while combined with dimensional reduction because a single extracted feature forms a model with less complex rules and has an impact on decreasing the detection performance. Thus, feature selection and feature extraction, especially dimensional reduction, can be used to select features, generate new features or find the most relevant feature combinations to be compatible with the algorithms used in machine learning models to increase their accuracy and efficiency.
