End-milling is considered one of the well-known cutting processes that machines surface using cutter tools that operate at relatively high speed. In order to manufacture mechanical parts, existing studies suggest that minimum roughness could be obtained if the radial and axial depth of the cut, the feed rate, and the spindle speed is adequately adjusted. This study considers the results of an experimental investigation conducted on 30 samples of aluminum alloy AL-6061 using a CNC machine and analyzes the relationship between independent input parameters, taken within well-defined ranges, that characterize the machine setting and output performance of surface roughness. Because the surface finish is a major concern, the present study considers an Adaptive Network-based Fuzzy Inference System (ANFIS) for the modeling: The paper presents results of surface roughness prediction using ANFIS. Emphasis is put upon shapes of membership functions, types of FIS generation, and FIS training optimum methods as these configurations affect the most performance of the ANFIS model. Results expressed in terms of an accuracy tool named root mean square error (RMSE) are compared. ANFIS has a strong learning capacity, prediction accuracy and can capture nonlinearity problems. In addition, The ANFIS surface roughness prediction model is expected to improve the cutting processes and the manufacturing process in the industry through sustainable prediction analysis.