Combining nearest neighbor data description and structural risk minimization for one-class classification

Abstract

One-class classification is an important problem with applications in several different areas such as novelty detection, anomaly detection, outlier detection and machine monitoring. In this paper, we propose two novel methods for one-class classification, referred to as NNDDSRM and kNNDDSRM. The methods are based on the principle of structural risk minimization and the nearest neighbor data description (NNDD) one-class classifier. Experiments carried out using both artificial and real-world datasets show that the proposed methods are able to significantly reduce the number of stored prototypes in comparison to NNDD. The experimental results also show that the proposed methods outperformed NNDD-in terms of the area under the receiver operating characteristic (ROC) curve-on four of the five datasets considered in the experiments and had a similar performance on the remaining one.

A Comparative Analysis of One-class Structural Risk Minimization by Support Vector Machines and Nearest Neighbor Rule

“…Finally, we compare the performance of the kernel k -NNDDSRM with our earlier method, the k -NNDDSRM [4]. Table 3 shows the best results obtained in this paper and in [4], considering the same data sets.…”

“…The first artificial data set was generated from two Gaussian Distributions and was also used in [3] [4]. In the Gaussian Distributions data set the samples belonging to normal class were generated by a Gaussian distribution with mean 0 and covariance 4 and the samples belonging to novel class by one with mean 4 and covariance 4.…”

“…In a recent paper, we proposed to implement one-class classification with the SRM principle using a nearest neighbor (NN) rule, referred to as k -NNDDSRM [4]. One of the objectives of k -NNDDSRM is to reduce the number of instances in an NNDD like one-class classifier while improving its classification performance.…”

“…One of the objectives of k -NNDDSRM is to reduce the number of instances in an NNDD like one-class classifier while improving its classification performance. Analysis has shown that this new method had a lower complexity in comparison with the NNDD [15] with an improved performance in almost all data sets considered in the experiments [3,4].…”

“…To evaluate the effectiveness of our proposed method we conducted some experiments using both artificial and real-world data sets and compared it with both the SVDD [15], Support Vector Data Description, and the original k -NNDDSRM [4]. In this paper we have chosen the SVDD by its SVM nature which means we are dealing with one of the more sophisticated and powerful methods available today.…”