In-process surface roughness (Ra) prediction systems for CNC machining incorporate quality control inspection while machining instead of performing inspection post-production. Such systems generally consist of a sensing technology and a decision making model. Data preprocessing is established as a necessary but often overlooked step in data analysis. Using acoustic emission (AE) signals as a sensing technology and artificial neural network (ANN) as a decision making model, the researchers were able to build and compare two ANN models using input variables feed rate, frequency, and peak volume— an original unfiltered ANN and a quality control-based filtered ANN which uses an individual quality control chart as a data preprocess. Based on a milling process dataset, results show that the QC-based ANN model makes better, more accurate Ra predictions, with a mean square error (MSE) of 0.0214, compared to the original unfiltered ANN, with an MSE of 0.0189. These results prove the success of the QC-based in-process ANN Ra prediction system.