<p>Chemical mechanical planarization (CMP) is widely used to planarize semiconductor surfaces in semiconductor manufacturing. However, in CMP, which involves various process variables, selecting the process conditions for the planarization of wafers is difficult without experience. Moreover, predicting CMP results depending on the usage time of consumables and various environmental factors is challenging. Therefore, this study attempted to predict the normalized material removal rate profile of a CMP machine with a five-zone carrier head using a deep neural network. Of a total of 80 experimental patterns, 67 were used for learning and 13 were used for validation, and the prediction accuracy was verified using five prediction patterns. In this study, the learning network showed an average prediction accuracy of 97.50% for the five prediction patterns.</p>