Determination of oxidant exposure values during the ozonation process is critical in order to measure the extent of micropollutant oxidation in the target water. While it is somewhat possible to determine ozone (O₃) exposure using in-situ sensors in large-scale water treatment plants, realistic hindrances make it challenging to accurately calculate hydroxyl radical (^•OH) exposure in the field. To solve this problem, prediction of ^•OH exposure using empirical modeling can be a viable option. In this study, instantaneous O₃ demand (IOD) and O₃ exposure were used as additional input parameters alongside basic water characteristic parameters and ozonation experiment conditions for ^•OH exposure prediction. User-defined response surface methodology (RSM) as well as various machine learning methods were used to develop models with different input parameter combinations for the prediction of ^•OH exposure values. Model training and validation results showed that models developed using O₃ exposure as an additional input parameter generally showed better prediction accuracy compared to models that did not. IOD was determined not to be as important as a potential input parameter when compared with O₃ exposure. Additional comparison of predictive abilities between machine learning methods as well as sensitivity analysis were conducted for further discussion.