RF-EKHO: Random forest with enhanced krill herd optimization algorithm for proficient detection of outliers in data with high-dimensions

Abstract

The detection of outliers is a challenging issue in the case of data with high dimensions. It is extensively used in distinct fields of study like social networks, knowledge discovery and statistics. To maintain the network privacy and security in social networks identify structural abnormalities in a constructive way, which are different from the typical behavior of the social network. In this paper, we propose a hybrid model to discover outliers in social networks utilizing Random Forest (RF) and Enhanced Krill Herd Optimization (EKHO) algorithm. The RF is used to enhance the execution and exactness of general procedure and it is a productive classification strategy. The leaves per tree and the trees per the forest are the two parameters of RF. Experimental results shows the efficiency and success of proposed method in terms of accuracy, detection rate, and computational time.