Dynamic rough-fuzzy support vector domain description for outlier detection

Abstract:

Outlier detection is one of the main data mining pre-processing tasks with real-world applications in areas like, e.g., fraud detection or medicine, where the 'rare cases' are of particular interest. Many techniques have been developed and used successfully in such situations. In changing environments, however, the existing outlier detection systems need to be regularly updated in order to describe in the best possible way an observed phenomenon at each point in time. Since changes lead to uncertainty, the respective systems also require an adequate modeling of the involved kinds of uncertainty. This paper presents a novel model for unsupervised dynamic outlier detection called Dynamic Rough-Fuzzy Support Vector Domain Description (D-RFSVDD). Its main idea is to update the outlier detection model in cycles of newly arriving observations considering simultaneously changes of past observations' feature values. This way, it takes advantage of the knowledge acquired in previous cycles to accelerate model updating while tracking the structural changes that the target and outlier classes experience over time. The core method of the proposed approach is the well-known Support Vector Domain Description algorithm which can be used for large data sets employing powerful optimization techniques. Our computational experiments highlight the potential D-RFSVDD has in dynamic environments.

Año de publicación:

2018

Keywords:

Fuente:

scopus