A QSPR study was performed between descriptors representing the molecular structures and the ozonolysis half-lives of azo dyes. Five descriptors were selected on the training set by stepwise regression. Based on these five descriptors, a model with the squared correlation R2 = 0.90 was developed by using SW-MLR. All descriptors involved in the model were derived solely from the chemical structures of the dyes, which makes the model very useful to estimate the ozonolysis half-lives of the dyes before they are synthesized. Classification functions were also developed using the linear discriminant analysis (LDA) to classify studied azo dyes into foaming or non-foaming compounds. The reliability of the proposed models was further illustrated using various evaluation techniques.
