Enhancing the feature selection by employing improved optimization with Simulated Annealing Algorithm for dimensionality reduction in intrusion detection dataset
Affiliations: [a] Department of Computing Technologies, School of Computing, SRM Institute of Science and Technology, Kattankulathur, Chennai, Tamilnadu, India
| [b] Department of Computer Science & Engineering (CS), School of Engineering, Malla Reddy University, Hyderabad, India
| [c] Department of Computer Science & Engineering, Vignan’s Foundation for Science, Technology & Research, Guntur, Andhra Pradesh, India
| [d] School of Computer Science and Artificial Intelligence, SR University, Warangal, Telangana | [e] Department of Artificial Intelligence and Data Science, Rajalakshmi Institute of Technology, Chennai, India
Correspondence:
[*]
Corresponding author. A. Arulmurugan, Department of Computing Technologies, School of Computing, SRM Institute of Science and Technology, Kattankulathur, Chennai, Tamilnadu, 603203, India. E-mail: [email protected].
Abstract: In the current scenario, feature selection (FS) remains one of the very important functions in machine learning. Decreasing the feature set (FSt) assists in enhancing the classifier’s accuracy. Because of the existence of a huge quantity of data within the dataset (DS), it remains a colossal procedure for choosing the requisite features out of the DS. Hence, for resolving this issue, a new Chaos Quasi-Oppositional-based Flamingo Search Algorithm with Simulated Annealing Algorithm (CQOFSASAA) has been proffered for FS and for choosing the optimum FSt out of the DSs, and, hence, this lessens the DS’ dimension. The FSA technique can be employed for selecting the optimal feature subset out of the DS. Generalized Ring Crossover has been as well embraced for selecting the very pertinent features out of the DS. Lastly, the Kernel Extreme Learning Machine (KELM) classifier authenticates the chosen features. This proffered paradigm’s execution has been tested by standard DSs and the results have been correlated with the rest of the paradigms. From the experimental results, it has been confirmed that this proffered CQOFSASAA attains 93.74% of accuracy, 92% of sensitivity, and 92.1% of specificity.
