Hellenic Open University Conferences, International Conference on Business & Economics of the Hellenic Open University 2015

Font Size: 
Using multiclass Support Vector Machines to forecast the occurrence of price spikes in the German electricity market
Efthymios Stathakis, Periklis Gogas, Theofilos Papadimitriou

Building: Titania
Room: Solon
Date: 2015-02-07 04:00 PM – 06:00 PM
Last modified: 2015-01-27


In this paper, employing high-frequency data from the German EPEXSpot electricity market we develop a model to forecast the occurrence of spikes (positive and negative) in the natural logarithmic returns of the hourly electricity prices. The analytical framework we use to model the extreme behavior of the intraday electricity prices is the Extreme Value Theory (EVT). The empirical model we employ to forecast the occurrence of spikes in the return series is the Support Vector Machines (SVMs).


Extrene Value Theory, Support Vector Machines, multiclass, forecast, EGARCH, model