Pearson Correlation Attribute Evaluation-based Feature Selection for Intrusion Detection System

IDS helps to overcome the network attack by taking appropriate preventive measures. The data mining method has good adaptability to new attack types; however, it consumes much time for high dimensional data. Therefore, the system needs a reduction of that high dimension. In this paper, we use a correlation approach of the attribute to evaluate those high dimensional data. To achieve a better environment, we propose a cut-off value of correlation to select some best features to use in the classification process. The best cut-off value in our experiment is 0.2 in RF classification that reaches 99.36% accuracy. The selection feature can reduce the time consumed in the running system.