Abstract:Capturing the dependence structure of multivariate extreme events is a major concern in many fields involving the management of risks stemming from multiple sources, e.g. portfolio monitoring, insurance, environmental risk management and anomaly detection. One convenient (non-parametric) characterization of extremal dependence in the framework of multivariate Extreme Value Theory (EVT) is the angular measure, which provides direct information about the probable 'directions' of extremes, that is, the relative contribution of each feature/coordinate of the 'largest' observations. Modeling the angular measure in high dimensional problems is a major challenge for the multivariate analysis of rare events. The present paper proposes a novel methodology aiming at exhibiting a sparsity pattern within the dependence structure of extremes. This is done by estimating the amount of mass spread by the angular measure on representative sets of directions, corresponding to specific sub-cones of $R^d\_+$. This dimension reduction technique paves the way towards scaling up existing multivariate EVT methods. Beyond a non-asymptotic study providing a theoretical validity framework for our method, we propose as a direct application a --first-- anomaly detection algorithm based on multivariate EVT. This algorithm builds a sparse 'normal profile' of extreme behaviours, to be confronted with new (possibly abnormal) extreme observations. Illustrative experimental results provide strong empirical evidence of the relevance of our approach.

Statistical inference on $D^{(d)}(u_n)$ condition and estimation of the Extremal Index

Clustering of extreme values: estimation and application

Estimation and Inference of Extremal Quantile Treatment Effects for Heavy-Tailed Distributions

Statistical Inference on a Changing Extremal Dependence Structure

A remarkable example on clustering of extremes for regularly-varying stochastic processes

On the spatial extent of extreme threshold exceedances

On estimation and order selection for multivariate extremes via clustering

Extremal clustering and cluster counting for spatial random fields

Extremes for stationary regularly varying random fields over arbitrary index sets

On the asymptotics of extremal lp-blocks cluster inference

Estimation of extreme risk regions under multivariate regular variation

Statistical inference for multivariate extremes via a geometric approach

Bayesian spatial clustering of extremal behaviour for hydrological variables

Inference on Extreme Quantiles of Unobserved Individual Heterogeneity

Estimating Extreme Value Index by Subsampling for Massive Datasets with Heavy-Tailed Distributions

Extremum statistics: A framework for data analysis

On the estimation of the extremal index based on scaling and resampling

Patterns in Spatio-Temporal Extremes

Sparsity in Multivariate Extremes with Applications to Anomaly Detection

On the Impact of Approximation Errors on Extreme Quantile Estimation with Applications to Functional Data Analysis

Bayesian Inference for Non-Parametric Extreme Value Theory