Outlier Detection Using the Information Entropy of Neighborhood Rough Sets

摘要

As an important research direction in data mining, outlier detection has been drawing much attention from different communities. Much use has been made of rough set theory for outlier detection as it is completely data-driven and no other knowledge is required; most other methods require some additional information. However, traditional rough set-based methods in the literature are restricted to the requirement that all data must be discrete. It is therefore not possible to consider real-valued or noisy data. This is usually addressed by employing a discrete method, which can result in information loss. This paper proposes a new approach based on the neighborhood rough set model, which has the ability to deal with real-valued data whilst simultaneously retaining dataset semantics. More significantly, this paper describes the underlying mechanism for this new approach to utilize the information entropy for measuring uncertainty. The use of this measurement can improved outlier detection accuracy. These results are supported by an experimental evaluation which compares the proposed approach with a number of existing outlier detection techniques.