The quantitative structure-activity relationship (QSAR) models were used to predict the photooxidation reaction of aromatic micro-pollutants (AMPs) using the multiple linear regressions (MLR) and support vector machine (SVM). The dataset includes of 30 compounds, divided into the training and test subsets by hierarchical clustering method. A genetic algorithm (GA) was employed to select the most relevant molecular descriptors, and multiple linear regression (MLR) and support vector machine (SVM) methods were subsequently used to establish linear and nonlinear models, respectively. The GA-MLR model based on three descriptors provided R2train=0.837, RMSEtrain=0.184, R2test=0.781, and RMSEtest=0.231. The nonlinear GA-SVM model showed improved predictive performance, with R2train=0.955, RMSEtrain=0.098, R2test=0.862, and RMSEtest=0.171. Cross-validation, Y-randomization, external hold-out evaluation, and applicability-domain analysis were used to assess model reliability. The selected descriptors, ESP-DPSA-1, SPH, and R3m+, indicate that electrostatic surface imbalance, molecular compactness, and mass-weighted three-dimensional atomic distribution contribute to the photooxidation behavior of AMPs. Within its defined chemical domain, the developed model may be used as a screening tool for prioritizing structurally related aromatic micropollutants for experimental photooxidation studies.