The paper evaluates the feasibility of monitoring cutting forces for in-process prediction of the workpiece surface roughness, using regression based models (RG) and artificial neural network (ANN) techniques. The three orthogonal cutting force components (F-x, F-y, F-z,) and the machined length L have been chosen as input variables. In the experimental test, AISI-1045 steel material was turned using a TiN coated carbide tool and employing a range of machining conditions (cutting speed: v=150, 200, 250 m/min; feed rate:f=0.15, 0.20, 0.25 mm/rev; depth-of-cut: d=l, 2, 3 mm). The results provided a wide range of measured cutting force and surface roughness values (Ra and Rq), which were used for adjustment and validation of the prediction models. Two prediction models were developed and subsequently the model accuracy was assessed by comparing the surface roughness predicted by the models with that measured by a 2D profilometer. The results highlighted the reasonably good fit given by both models, with the ANNbased model providing best accuracy for surface roughness prediction. The prediction of the output surface roughness in an automated turning process was established and was found to be feasible by the monitoring of cutting forces.