Feature Selection Method Based on Correlation Tree

Machine learning is one of techniques adapted to detect intrusion for cyber security. One of importance techniques to find anomaly is classification. But classification with huge dataset has the resources and time consumption. Feature selection is choice to reduce the data dimension to improve processing performance. In this paper, we introduce the new feature selection method that selects some fields of data set using position of each feature in correlation tree. Then, the result from the correlation tree feature selection of KDDCUP’99 data set are compared with two feature selection technique, correlation of coefficient (CC-type) and BFS by using three reference classifier, Decision Tree (DT), Random Forest (RF), and Naive Bayes (NB).