Abstract: From MITI-Ⅰ test,a total of 587 different kinds of organic compounds were screened out.Through structure splitting,50 kinds of these were picked out randomly as a validation set,and the other 537 as a training set.Models were built up using the multiple linear regression method(MLR) and support vector machine(SVM),separately,for predicting aerobic biodegradation of organic chemicals.Results show that functional groups,such as aromatic acid,aldehyde,aromatic iodine and tetriary amine,had much effect on aerobic biodegradability of these organic compounds.The prediction using the multiple-linear-regression-based model reached 81.43% in overall accuracy,and 82% with the validation set;and the use of the other model based on support vector machine reached 87.90% in overall accuracy,and 86% with the validation set.The two kinds of QSBR models can be used to evaluate aerobic biodegradability of organic chemicals. 

Key words: aerobic biodegradation, multiple linear regression (MLR), support vector machine (SVM), quantitative structure-activity relationship (QSAR)