Outlier detection method based on hybrid rough

Abstract

This paper discusses on the detection of outliers by hybriding Rough_Outlier Algorithm with Negative Association Rules. An optimization algorithm named Binary Particle Swarm Optimization is used to improve the computation of Non_Reduct in order to detect outliers. By using Binary PSO algorithm, the rules generated from Rough_Outliers algorithm is optimized, giving significant outliers object detected. The detection of outliers process is then enhanced by hybriding it with Negative Association Rules. Frequent and Infrequent item sets from outlier rules are generated. Results show that the hybrid Rough_Negative algorithm is able to uncover meaningful knowledge of outliers from the frequent and infrequent item sets. These knowledge can then be used by experts in their field of domain for better decision making.