Combining a class-based classification method with ensemble boosting and sequential feature selection

Abstract

The Binary Min-Redundancy Max-Diversity (BMRMD) was utilized to determine the computer network hacking and attacks. The Intrusion Detection System (IDS) is crucial for detecting attacks on an organization, which have increased in size and scale, as well as other anomalies. IDS achieves this by preparing for the unauthorized information related to network security and it is essential for distinguishing various types of attacks. The organization's traffic dataset contains numerous highlights, so selecting and eliminating irrelevant items improves the accuracy of the organization's calculations. Containing a large amount of meaningless or excessive data, a dataset can cause fitting problems and reduce the capacity of the model to learn meaningful patterns. BRMMD approach covers not only the significance of each element but also the expected accuracy when an ideal set of features is given. Solving such challenges requires a series of feature selection techniques. Therefore, the challenge is addressed by evaluating the repeatability of the features and determining their relevance to the target class based on the optimal grouping of the included features.