Surface ozone (O₃) is a major atmospheric pollutant that negatively impacts both human health and ecosystems. This research aimed to develop a combined model (PC-ANN) for predicting surface O₃ concentrations using Principal Component Analysis (PCA) and Artificial Neural Network (ANN) models in Baghdad. The input data included meteorological variables such as temperature (T2m), total precipitation (TP), boundary layer height (BLH), surface pressure (SP), wind components (V10m, U10m), along with pollutants like NO₂, CO, SO₂, PM2.5, and PM10. The findings indicated significant variations in the correlation between O₃ levels and the selected variables. The dataset's suitability for factor analysis was confirmed through KMO and Bartlett's tests, with a KMO value of 0.836 for Baghdad, and Bartlett’s test resulting in a p-value ˂ 0.001(0.000). After applying varimax rotation, three principal components (PCs) were identified for Baghdad, accounting for 71.827% of the total variance. PC1, PC2, and PC3 were primarily linked to NO₂, PM2.5, and T2m, respectively. For the ANN model, various network structures were tested, and the optimal performance This was achieved using 4 nodes in the hidden layer and the tanh activation function resulting in an adjusted R2 value of 0.871 for Baghdad. According to the ANN model, the variables derived from PCA—FAC-1 (NO₂), FAC-3 (T2m), and FAC-2 (PM2.5)—had importance scores of 100.0%, 74.6%, and 6.65%, corresponding to 0.552, 0.412, and 0.037, respectively. Additionally, the PC-ANN model effectively captured nonlinear relationships, demonstrating stronger predictive performance for complex datasets, with higher R2 and lower RMSE and MAE values.