The potential of using three different data-driven\ntechniques namely, multilayer perceptron with backpropagation\nartificial neural network (MLP), M5 decision\ntree model, and Takagiââ?¬â??Sugeno (TS) inference system for\nmimic stageââ?¬â??discharge relationship at Gharraf River system,\nsouthern Iraq has been investigated and discussed in\nthis study. The study used the available stage and discharge\ndata for predicting discharge using different combinations\nof stage, antecedent stages, and antecedent discharge values.\nThe modelsââ?¬â?¢ results were compared using root mean\nsquared error (RMSE) and coefficient of determination (R2)\nerror statistics. The results of the comparison in testing\nstage reveal that M5 and Takagiââ?¬â??Sugeno techniques have\ncertain advantages for setting up stageââ?¬â??discharge than\nmultilayer perceptron artificial neural network. Although\nthe performance of TS inference system was very close to\nthat for M5 model in terms of R2, the M5 method has the\nlowest RMSE (8.10 m3/s). The study implies that both M5\nand TS inference systems are promising tool for identifying\nstageââ?¬â??discharge relationship in the study area.
Loading....