In this paper the photolysis half-lives of the model dyes in water solutions and under ultraviolet (UV) radiation were determined by using a continuous-flow spectrophotometric method. A quantitative structure-property relationship (QSPR) study was carried out using 21 descriptors based on different chemometric tools including stepwise multiple linear regression (MLR) and partial least squares (PLS) for the prediction of the photolysis half-life (t_1/2) of dyes. For the selection of test set compounds, a K-means clustering technique was used to classify the entire data set, so that all clusters were properly represented in both training and test sets. The QSPR results obtained with these models show that in MLR-derived model, photolysis half-lives of dyes depended strongly on energy of the highest occupied molecular orbital (E_HOMO), largest electron density of an atom in the molecule (ED⁺) and lipophilicity (logP). While in the model derived from PLS, besides aforementioned E_HOMO and ED⁺ descriptors, the molecular surface area (S_m), molecular weight (MW), electronegativity (χ), energy of the second highest occupied molecular orbital (E_HOMO-1) and dipole moment (μ) had dominant effects on logt_1/2 values of dyes. These were applicable for all classes of studied dyes (including monoazo, disazo, oxazine, sulfonephthaleins and derivatives of fluorescein). The results were also assessed for their consistency with findings from other similar studies.