Abstract

Outlier detection is a fundamental issue in data mining, specifically it has been used to detect and remove anomalous objects from data. It is an extremely important task in a wide variety of application domains. The existing research method named as Expectation Maximization Particle Swarm Optimization Weighted Clustering (EMPWC) outlier detection technique is used for detecting the outliers more efficiently. However it has issue with handling the unbalanced dataset and time complexity. To avoid the above mentioned issues, in this research, BAT optimization based semi supervised algorithm is proposed. It has three phases such as pre-processing, outlier estimation using BAT algorithm and clustering using semi-supervised algorithm. The pre-processing is done by using min-max normalization approach which is sued to increase the outlier detection accuracy. The outlier detection is improved by using BAT optimization algorithm which is achieved by best objective function. The clustering is done by Graph based Semi Supervised (GSS) algorithm. The GSS based BAT (GAABAT) approach is used to improve the performance metrics such as execution time and false alarm rate compare than existing methods. The experiment results on large scale categorical datasets have shown that the IBAT with semi supervised based outlier detection ensures a better trade-off between Detection Rate (DR), False Alarm Rate (FAR) than the existing outlier detection schemes.

Paper Details