Outlier detection via a soft computing hybrid

Outlier detection has been attracting the data analysts in almost every domain for a long time now because their detection is very challenging. Outliers or novel cases need to be detected before any analysis is performed on data set. Depending upon the domain, outlier detection saves a lot of time, money or both. In this paper, we developed a novel outlier detection model using ensembling technique, in the paradigm of soft computing, which includes four algorithms, namely k-Reverse Nearest Neighbor (kRNN), Auto Associative Neural Network (AANN), Counter Propagation Auto Association Neural Network (CPAANN), and Generalized Regression Auto Association Neural network (GRAANN) as constituents. The ensemble takes the union of all the outliers found by the four techniques.