In this study, we explore the process performance of flank-end milling of Ti-6Al-4V titanium alloy. Experiments and convolutional neural networks are used to establish a predictive model of machining quality. Sensory tool holders are used to capture the cutting force signals during machining and to perform feature extraction. The neural network model utilizes feature data as input with surface roughness and dimensional accuracy as outputs. The experimental framework can be divided into several stages: machining, cutting data collection, surface roughness and machining accuracy measurement, and neural network parameter setting. The experimental parameters consisted of cutting speed, feed per tooth, axial cutting depth, and radial cutting depth. Each parameter has three levels. Therefore, for a full-factor experiment, 81 sets of experimental data are obtained. Furthermore, 162 sets of data are obtained by performing each experiment twice. In the neural network prediction results, the minimum average percentage for surface roughness prediction error is below 10% when grouping the feed per tooth. This result was considered favorable compared with the error percentage of 18% obtained from predictions through training with all data. On the other hand, the machining accuracy prediction results were better when training with all data, with the error percentage being approximately 20%.