The semiconductor industry is always looking for new solutions to maximize yield. Recently, the focus has been on utilizing the manufacturing data to help improve operational efficiency and early detection. This paper proposes a framework to find the best combination of machine learning models and data-balancing methods to predict specific wafer map signatures using Wafer Acceptance Test (WAT). WAT is a measurement test performed at multiple locations to identify poorly manufactured wafers. However, there were instances where wafers passed every measurement test but were found to have low yield. The proposed framework will be tested on real manufacturing data to demonstrate the viability of predicting wafer map signatures.