**3.2 M5 model tree**

The outcomes of training and testing of the M5 Model Tree for the rainfallrunoff process confirms the fact that it has the potential of identifying the relationship between both hydrological variables of a catchment. This statement was confirmed by the model evaluation criteria with low values of NRMSE and high values of R<sup>2</sup> and COE for the validation and testing of the dataset, which suggests the best model fit. The visualization of **Table 4** shows that the M5 Model Tree has the capability to reproduced well by the model with different rainfall-runoff input combinations. The training results indicate that the prediction of Q(t-2) and Q(t-3) quite well for the rainfall-runoff process having results of R<sup>2</sup> , COE, MSE and NRMSE, 0.71, 1.00, 0.00, 757158.18 m<sup>3</sup> /sec and 0.71, 1.00, 0.00, 757158.18 m<sup>3</sup> /sec respectively. During testing of the model, the model evaluation parameters R2 , COE, MSE and NRMSE results are found as 1.00, 1.00, 0.00, 887.52 m<sup>3</sup> /sec with input C7, which means that the M5 model tree explicit good results in testing with both rainfall and runoff combinations. The modeling error for the verification of the results indicates high values of R2 and COE and low values of NRMSE, demonstrating the good M5 model tree performance.
