A Novel Outlier Detection Method for Multivariate Data

Detecting anomalous objects from given data has a broad range of real-world applications. Although there is rich number outlier detection algorithms, most them involve hidden assumptions and restrictions. This paper proposes novel, yet effective learning algorithm that based on decomposing the full attributes space into different combinations subspaces, in which 3D-vectors, representing points per 3D-subspace, are rotated about geometric median, using Rodrigues rotation formula, to construct overall outlying score. The proposed approach parameter-free, requires no distribution easy implement. Extensive experimental study comparison conducted both synthetic datasets with six popular each category. evaluated precision @s , average precision, rank power, AUC ROC time complexity metrics. results show performance method competitive promising.