Finding Anomaly with Fuzzy Rough C-Means Using Semi-Supervised Approach

Outlier detection is initial step in various data-mining applications. This methods have been suggested for number of applications, such as credit card fraud detection, clinical trials, voting irregularity analysis, data cleansing, network intrusion, severe weather prediction, geographic information systems, athlete performance analysis, and other data-mining tasks proposed algorithm. In this proposed system combines the fuzzy set theory, rough set theory and semi-supervised learning to detect outliers and is a new try in area of outlier detection for semi-supervised learning. Without considering those points located in lower approximation of a cluster, proposed algorithm only need discuss the possibility of the points in boundary to be assigned as outliers and has many advantages over SSOD. proposed algorithm uses labelled normal and outliers and as well as samples without labels and can improve outliers detection accuracy and reduce false alarm rate under the guidance of labeled samples. proposed algorithm will be applied to many outlier detection fields which has only partially labeled samples, especially that does not make a certain judgment in uncertain conditions. But, the results depend on selection of number of cluster c, initial canter of cluster, parameters, proposed algorithm usually also stops on a local minimum. So, during the process, It must carefully select initial canters and other parameters. The proposed system proposes the technique that may add parameters to speed up the technique.