Flower Pollination Algorithm for feature analysis of Kyoto 2006+ data set

Abstract

Unwanted and large features in data contributes to network classification problem. Features can be selected to improve the performance. In this paper, Flower Pollination Algorithm (FPA) has been proposed to select the optimal features. Further, three predefined feature selection algorithms have been used to selects the most critical attributes for anomaly detection. The performance of FPA and three predefined algorithms have been compared on fifteen features of kyoto 2006+ of Intrusion Detection System (IDS). The evaluation results show that ‘Service (2), Srv_serror_rate (8) and Flag (14) features are the most critical features.