One of the important issues in water transport and sewer systems is determining the flow resistance and roughness coefficient. An accurate estimation of the roughness coefficient is a substantial issue in the design and operation of hydraulic structures such as sewer pipes, the calculation of water depth and flow velocity, and the accurate characterization of energy losses. The current study, applies two kel based approaches [Support Vector Machine (SVM) and Gaussian Process Regression (GPR)] to develop roughness coefficient models for sewer pipes. In the modeling process, two types of sewer bed condi-tions were considered:loose bed and rigid bed. In order to develop the models, different input combi-nations were considered under three scenarios (Scenario 1:based on hydraulic characteristics, Scenarios 2 and 3: based on hydraulic and sediment characteristics with and without considering sediment con-centration as input). The results proved the capability of the kel based approaches in prediction of the roughness coefficient and it was found that for prediction of this parameter in sewer pipes Scenario 3 performed better than Scenarios 1 and 2. Also, the sensitivity analysis results showed that Dgr (Dimensionless particle number) for a rigid bed and wb/y (ratio of deposited bed width, wb, to flow depth, y) for a loose bed had the most significant impact on the modeling process.