DefinePK

Abstract

The main objective of this paper is to determine the optimum values of the main parameters of Decision Tree Forest (DTF) ( Maximum Trees in Decision Tree Forest (MTDTF) , Minimum Size Node to Split (MSNS) , and Maximum tree levels (MTL) ) , and assessing of their effect on predicted average monthly discharge of Euphrates River, in Thi Qar province, southern Iraq. Four popular statistical parameters were used as evaluation criteria for evaluating DTF models performance: Coefficient of Correlation (R), Root Mean Squared Error (RMSE), Maximum Error (ME), and Mean Absolute Percentage Error (MAPE). Five-fold cross-validation is applied in this research to evaluate the performance of the DTF models. In the first stage of this study, the observed data volume involved in DTF models was equal to 240 months; in this stage the results illustrated that the optimum values of both MTDTF, and MSNS were equal to 125 and 32 respectively, while the modification of the third parameter MTL does not show any effect on the performance of DTF, it was observed that the effect of this parameter on statistical parameters in the form of a straight line. In the second stage of this study, the observed data volume was increased to 480 months ; thus leading to increase the optimum value of MTDTF to 225 , and decrease the optimum value of MSNS to 2 ,while the results does not show any sensitivity to the parameter MTL.