QSBR Models for Predicting Anaerobic Biodegradation of Chemicals

Anaerobic biodegradation data of 155 kinds of organic chemicals were collected.Among them the data of 109 chemicals were picked up randomly for use in the training set,and of the remaining 46 in the validation set.By splitting constitutional formula,radicals or groups of the chemicals were obtained and analyzed with multiple linear regression and BP artificial neural network,separately,to explore for quantitative relationships between structure and biodegradability of organic chemicals(QSBR).Results show that the use of MLR in analyzing the validation set was 78.26% in accuracy and 84.52% in total accuracy and for the use of BP-ANN,82.61% in accuracy and 90.32% in total accuracy.Obviously the latter is superior to the former.