Abstarct: The industrial textile effluents contain a large variety of dyes that without proper treatment prior to their disposal become toxic to microorganisms and aquatic life. The indiscriminate release of dyes has a considerable negative impact on the environment and human health. In this study, QSPR model baxsed on Bayesian regularized artificial neural network (BR-ANN) was developed for the modelling and accurate prediction of photo-catalytic dye decolorization efficiency at differentiate baxsed on their chemical structure. For this purpose, the photo-catalytic decolorization of various dyes (31 dyes) have been investigated in TiO2/UV suspensions. Under constant experimental conditions including amount of photo-catalyst (0.0100g), sample volume (50.0 ml) and pH of 7.0 decolorization efficiency of each dye were measured at 5, 10 and 15 minutes after initiation of UV irradiation. Dyes structures were converted to the molecular descxriptors using Dragon software and the most significant descxriptors were selected and used as inputs in ANN modelling. After obtaining the BR-ANN model as an optimum model, the prediction ability of the model was evaluated by test set and leave-one-out method. The squared correlation coefficients obtained for the test set and LOO were 0.9553 and 0.9320, respectively. The results obtained, showed the superior prediction ability of the proposed model in the prediction of decolorization efficiency.