A new application of principal component analysis- wavelet transform- artificial neural network (PC-WT-ANN) was proposed for the simultaneous determination of uranium and thorium. A simple spectrophotometric method for simultaneous determination of uranium and thorium based on formation of their complexes with 4- (2-pyridylazo) resorcinol (PAR) is described. Although the complexes of uranium and thorium with reagent show a spectral overlap, they have been simultaneously determined by PC-ANN and PC-WT-ANN. The results obtained by the two methods were compared and it was shown that in PC-WT-ANN, the convergence speed was faster and the root mean square error of prediction (RMSEP) was also smaller than PC-ANN. The results also show very good agreement between true and predicted concentration values and have the ability to use in routine analysis.