Approximate polytope ensemble for one-class classification

Abstract

a b s t r a c tIn this work, a new one-class classification ensemble strategy called approximate polytope ensemble is presented. The main contribution of the paper is threefold. First, the geometrical concept of convex hull is used to define the boundary of the target class defining the problem. Expansions and contractions of this geometrical structure are introduced in order to avoid over-fitting. Second, the decision whether a point belongs to the convex hull model in high dimensional spaces is approximated by means of random projections and an ensemble decision process. Finally, a tiling strategy is proposed in order to model nonconvex structures. Experimental results show that the proposed strategy is significantly better than state of the art one-class classification methods on over 200 datasets.

Detección de anomalías basada en técnicas inteligentes de una planta de obtención de material bicomponente empleado en la fabricación de palas de aerogenerador

“…Esta técnica de clasificación one-class (tipo 1) se ha revelado como una herramienta efectiva a la hora de detectar anomalías en diversas disciplinas (Casale et al, 2011;Fernández-Francos et al, 2018). Además, ofrece resultados satisfactorios en relación con otras técnicas one-class aplicadas a conjuntos de datos de repositorios UCI (Universidad de California en Irvine) (Casale et al, 2014). La idea principal de esta técnica de clasificación consiste en modelar el límite de un conjunto de datos D ∈ R n a partir su contorno convexo.…”

Deep Learning Based Surveillance System for Open Critical Areas

“…We consider a fully supervised approach to tackle anomaly detection and localization as a one-class classification problem. We use the Polytope Ensemble technique [28] as the modeling method. This method represents an approximation of the space containing the input feature samples with a set of polytopes.…”

“…As a result, we increase the robustness of the model by modifying the structure of the convex hull. We shift the vertices of the convex hull farther or closer to the centroid, similarly to [28]. In detail, given the centroid of the polytope c i and its set of vertices V ⊂ X, we compute an expanded polytope by correcting the vertices such that…”

Ensemble of One-Class Classifiers for Personal Risk Detection Based on Wearable Sensor Data

“…Additionally, a promising approach for enhancing an ensemble involves training one or several classifiers of the ensemble with different features of the dataset, where the most common techniques are random feature selection methods, such as the random subspace method (RSM, which is also called attribute bagging) and random partitions, subspace clustering, projection methods, bagging and rotation forest [2,3,4,5,6,7,8,9,10,11,12,13]. …”