B1163
Title: An algorithmic method for fitting multimodes heavy-tailed distributions
Authors: Marie Kratz - ESSEC Business School, CREAR (France) [presenting]
Abstract: A hybrid model is proposed for heavy-tailed phenomena, combining a Gaussian Mixture Model (GMM) with a Generalized Pareto component (GPD). It generalizes a previous algorithmic method, with an automatic detection of the tail threshold. While the main advantage of the GMM is the flexibility it affords when dealing with multimodal distribution, introducing a GPD component allows one to evaluate the tail of the distribution, unlike a pure GMM.
