An approach to develop a hybrid algorithm based on support vector machine and Naive Bayes for anomaly detection

摘要

Anomaly detection involves way towards finding the example in the information that violates ordinary conduct. The choice of anomaly detection algorithm can to a great extent affect the undertaking of anomaly identification. The decision of abnormality revelation calculation can influence complexity and correctness of the process. The choice of anomaly recognition calculations may increase the occurrence of false alert rate, high resource usage, and may even lead to security vulnerabilities. In addition, one anomaly detection procedure can beat the other in same dataset. In this way, many anomaly detection systems can be used to merge the prediction from multiple system in order to improve the generalizability over a single estimator. In this research work, we show a weighted hybrid model utilizing Support Vector Machine and Naive Bayes for anomaly discovery, k-fold cross validation to figure the error related with corresponding model and accuracy based weight values to be used with the candidate model. The hybrid algorithm has been executed to join the result of expectation of SVM and Naive Bayes classifiers utilizing weight elements. The weights elements have been computed utilizing root mean square error of forecast as error metric. The classifier with high accuracy has been given higher weight and classifier with the lower precision has been given lower weight. The objective is to improve the performance of hybrid model than that of Support Vector Machine (SVM) and Naive Bayes.