Efficient Prediction of Anomaly Detection Using Machine Learning Algorithms

Abstract

               In recent days, most threaten issue is Anomalous detections that has to be recognized as early as possible in order to avoid the unexpected suspicious observations. So, anonymous rule generation process should be developed to accomplish this task for early prediction of this issue from the database. The existing work Rule Pruning based Anomalous Rule Detection strategy (RPARD), concentrated much on this issue and attained its best. Here, the rule pruning is done through the stepwise regression analysis method. Special attention has to be given for the inconsistencies and inherent problem of multiple hypotheses testing in Stepwise regression. So, in our proposed work Lasso Regression based Improved Anomalous Detection Scheme (LR-IADS) was given to rectify and assure the optimal unexpected suspicious detection. Here, fuzzy anomalous rule were created initially with the help of attributes that are chosen from the data base. On the basis of Gini index, information gain and gain ratio, we choose the attributes here, and lasso regression analysis method helps to do the rule pruning on the generated anomalous rules. At last, unexpected suspicious detection is done according to these anomalous rules by commencing the classification process and it is performed by Enhanced Relevant Vector Machine based Association Classifier. Through MATLAB simulation environment, the simulation process is done and it is confirmed from the experiment that our proposed work LR-IADS assures better prediction result when compared with the earlier RPARD method.