The Analysis of the Numerical Price Forecasting Success Considering the Modification of the Initial Condition Value by the Commodity Stock Exchanges

Abstract

In mathematical models, for forecasting prices on commodity exchanges different mathematical methods are used. In the given paper the numerical model based on the exponential approximation of commodity stock exchanges was derived. The price prognoses of aluminium on the London Metal Exchange were determined as numerical solution of the Cauchy initial problem for the 1st order ordinary differential equation. To make the numerical model more accurate the idea of the modification of the initial condition value by the stock exchange was realized. The derived numerical model was observed to determine the success of the proposed modification by means of the distribution of the numbers of the forecasting terms with different error rate. Within the forecasting terms, in which the modification of the initial condition value by stock exchange occurred, the proposed strategy significantly improved the original forecasting. The largest forecasting problems within significant and rapid changes in price course were eliminated. The modification of the original model improved forecasting in each forecasting term where the initial condition drift occurred by reducing both the mean absolute percentage error of the forecasting term and the number of the prognoses with the absolute percentage error of at least 10 %.