Performance Analysis of Anomalous Detection Scehmes Based on Modified Support Vector Machine and Enhanced Relevance Vector Machine

Anomalous transactions are common activity happening on the financial oriented transaction. Detecting those anomalous transactions from the financial transaction patterns is the most complex task which is focused in this work. In the existing work it is achieved by introducing the method namely Fuzzy Exception and Fuzzy Anomalous Rule (FEFAR). The accuracy of this existing work FEFAR found to be lesser which is resolved in the proposed research work. There are two research works has been proposed those are namely Rule Pruning based Anomalous Rule Detection Strategy (RPARD) and Lasso Regression based Improved Anomalous Detection Scheme (LR-IADS). Both of these methods attempt to find the anomalous transaction from the given input database by finding the anomalous rules. Each method differ in its methodologies, thus the accuracy of the methods would differ. The main goal of this analysis work is to compare the performance of existing and proposed methodologies based on simulation outcome. This research work aims to highlight the performance variation between the proposed and existing techniques and the best method that can offer accurate anomalous transaction detection. The analysis of the research work is carried out on matlab environment over four databases namely soil, bank, german statlog and auto mpg based on which performance outcome has been given.