Chlorination of natural organic matters (NOM) in water could produce disinfection byproducts (DBPs) that are harmful to human bodies. In this study, we used ultra-high resolution mass spectrometry to identify the molecular formulas of NOM precursors of DBP after different water treatment processes (e.g., UV irradiation, activated carbon adsorption) and also to determine the changes in the molecular properties of NOM and their impacts to the DBP formation. A machine learning (ML) model was developed to predict the number of DBP formula from the molecular properties of NOM. A comparison of models using Pycaret showed that the catboost model had high prediction accuracy, suggesting that the molecular weight and carbon oxygen ratio are factors that have a significant impact on DBP formation. The results suggest that changes in the molecular properties of NOM, such as saturation degree and aromaticity (due to the UV irradiation treatment) may facilitate the number of DBP molecular formula formed.