RBDT: The cascading of machine learning classifiers for anomaly detection with case study of two datasets

The inhuman cause of behavior in computer users, lack of coding skills pursue a malfunctioning of applications creating security breaches and vulnerable to every use of online transaction today. The anomaly detection is in-sighted into security of information in early stage of 1980, but still we have potential abnormalities in real time critical applications and unable to model online, real world behavior. The anomalies are pinpointed by conventional algorithms was very poor and false positive rate (FPR) is increased. So, in this context better use the adorned machine learning techniques to improve the performance of an anomaly detection system (ADS). In this paper we have given a new classifier called rule based decision tree (RBDT), it is a cascading of C4.5 and Naïve Bayes use the conjunction of C4.5 and Naïve Bayes rules towards a new machine learning classifier to ensure that to improve in results. Here two case studies used in experimental work, one taken from UCI machine learning repository and other one is real bank dataset, finally comparison analysis is given by applying datasets to the decision trees (ID3, CHAID, C4.5, Improved C4.5, C4.5 Rule), Neural Networks, Naïve Bayes and RBDT. © Springer International Publishing Switzerland 2015.