A feature probability estimation-based feature selection approach for intrusion detection

This work proposes a probabilistic feature estimation-based feature selection method for intrusion detection systems. The method aims to reduce the number of insignificant features from the training set, which helps decrease the processing complexity and storage requirements during model training. It efficiently removes redundant or insignificant features and makes the system faster and more reliable. The proposed method applies to datasets that contain both data types: numerical and categorical data. We have compared the method with several well-known feature selection techniques: Principal Component Analysis (PCA), K-best (Chi-Square) feature selection method, Feature Importance method, and Recursive Feature Elimination (RFE) method. The experimental results show that the proposed method also shows a similar performance to existing techniques. The proposed method easily deals with large and complex datasets.