Energy commodity prices are a crucial variable in the economic context given their role in the consumption and production process, since energy commodities have recently become an asset class playing an increasing role in the risk management field. The pricing models and the techniques used to provide an estimation of price dynamics become more and more important. We propose in this paper a new methodology based on neural networks in order to build a forecast for specific standard price time series of energy commodities. In particular we analyze Crude oil, natural gas and electricity prices for the European and the US market. Using daily data over the period 2001-2010 we are able to provide very robust forecasts for a one year time horizon of the considered series. Furthermore, we perform a statistical assessment of the considered prediction models and we prove that some of them provide the first four unconditional moments of the predicted sequences almost equal to the moments estimated on the market data.
Neural Networks to Model Energy Commodity Price Dynamics / Panella, Massimo; Barcellona, Francesco; V., Santucci; D'Ecclesia, RITA LAURA. - ELETTRONICO. - (2011), pp. 1-4. (Intervento presentato al convegno 30th USAEE/IAEE North American Conference tenutosi a Washington D.C., U.S.A. nel 9-12 ottobre 2011).
Neural Networks to Model Energy Commodity Price Dynamics
PANELLA, Massimo;BARCELLONA, FRANCESCO;D'ECCLESIA, RITA LAURA
2011
Abstract
Energy commodity prices are a crucial variable in the economic context given their role in the consumption and production process, since energy commodities have recently become an asset class playing an increasing role in the risk management field. The pricing models and the techniques used to provide an estimation of price dynamics become more and more important. We propose in this paper a new methodology based on neural networks in order to build a forecast for specific standard price time series of energy commodities. In particular we analyze Crude oil, natural gas and electricity prices for the European and the US market. Using daily data over the period 2001-2010 we are able to provide very robust forecasts for a one year time horizon of the considered series. Furthermore, we perform a statistical assessment of the considered prediction models and we prove that some of them provide the first four unconditional moments of the predicted sequences almost equal to the moments estimated on the market data.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.