This work presents a new approach to predict thermal stability of organic peroxides based on quantitative structure-property relationship methods. The data set consists of self-accelerating decomposition temperature (SADT) of 39 organic peroxides, part of which is obtained from the Recommendations on the Transport of Dangerous Goods and the other part is obtained from C600 calorimetry. Descriptors computed using RM1 semi-empirical quantum mechanical method are selected to describe the above-mentioned 39 molecules. Partial least-squares method and multiple linear regression method are used to select the descriptors and SADT model. At last, a four-descriptor correlation equation is obtained using a cross validation method: a correlation coefficient R>0.9 and a predictive coefficient Q(2)>0.85. (C) 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.