UNSUPERVISED-ANOMALY-DETECTION

Unsupervised machine learning methods for novel anomaly detection.

CLASSICAL SVDD | code | KERNEL SVDD CODE | Paper

Support vector data description (SVDD) is an algorithm that defines the smallest hypersphere that contains all observation used for outlier detection or classification. Support Vector Data Description (SVDD) is also a variant of Support Vector Machines (SVM), usually referred to as the One class SVM. It is interesting for use cases where researchers are only interested in the positive class class of interest, therefore making it suitable to detect novel data or outliers.

CLASSICAL TWO CLASS SVDD | KERNEL SVDD CODE

In real life the target data set often contains more than one class of objects and each class of objects need to be described and distinguished simultaneously. Two class SVDD is an improved support vector data description method proposed to solve two class problem for novelty anomaly detection.