Metaheuristic Algorithms-based Feature Selection Approach for Intrusion Detection

Abstract

Since data mining is intended to extract the required knowledge from very large amounts of data, the presence of irrelevant or redundant attributes leads to poor predictability and increases the computation time. The concept of feature selection is considered one of the most important processes of data mining and machine-learning fields. Feature selection is the process of finding the least possible attributes that are able to describe the dataset as the original features do; this in turn, improves the performance of the prediction process and provides a better understanding of stored data. The success of selection process is based on the balance between two important factors, namely, the minimal number of reduced features and the high accuracy results. Feature-selection process is a complex optimization problem for which the optimal solution cannot be guaranteed, but it tries to find the best possible solutions. Therefore, feature-selection problems have been the focus of researchers in recent years, especially with the large increase in databases. Researches have been included to optimize feature selection problems using metaheuristic algorithms with learning model as one of the effective methods used in this field. In this chapter, many metaheuristic algorithms are applied in feature-selection problems. However, these algorithms possess the ability to deal with very complex problems, but the results depend on the nature and extent of the problem. Many complex problems can be solved by feature selection and data-mining techniques, like intrusion detection through mining network traffic, anomaly detection and data stream mining.