Forecasting of commodities prices using a multi-factor PDE model and Kalman filtering

Abstract

This study proposes a method for forecasting commodities prices using Schwartz partial differential equation (PDE) and Kalman filtering. The method is applicable to both the single-factor and the multi-factor Schwartz PDE. Using semi-discretisation and the finite differences method, the Schwartz PDE is transformed into an equivalent state-space description. This latter representation is finally written in a linear matrix form in which the Kalman filter's recursion is applicable. By redesigning the Kalman filter as a m-step ahead predictor it becomes possible to obtain accurate estimates of the future commodities’ price. The prediction scheme analysed in this study can contribute to maximising profits in commodities trading, including also the trading of electric power.

Keywords

• Kalman filters
• partial differential equations
• state-space methods
• finite difference methods
• pricing
• commodity trading
• matrix algebra
• profitability
• forecasting theory
• Kalman filtering
• multifactor Schwartz PDE
• finite differences method
• commodities trading
• multifactor PDE model
• Schwartz partial differential equation
• state-space description
• commodity prices forecasting
• single-factor Schwartz PDE
• semidiscretisation
• linear matrix
• m-step ahead predictor
• profit maximization