RDOD: A Robust Distance-based Technique for Outlier Detection
Reza Heydari Gharaei, Hossein Nezamabadi–pour · 2022
Outlier detection is an important topic in data mining and has been employed in various sciences. Numerous outlier detection methods have been proposed so far; one of the most prominent categories of these methods is based on k-nearest neighbor (k NN). In the present study, an efficient and robust distance-based outlier detection method is presented. One of the main challenges of methods based on k-nearest neighbor is their high dependency on parameter k. The proposed method, however, reduces the sensitivity to k while maintaining the high preciseness of the algorithm. The proposed method was evaluated in two-dimensional synthetic and multidimensional real datasets and compared with some state-of-the-art algorithms in the field. The results of the experiments proved the effectiveness of the proposed method.