Abstract: The basic principles of neural network and quantitative structure-property relationship（QSPR）study are introduced，and its application and advance in the prediction of flammability characteristics of compounds such as flash point，auto-ignition temperature，and flammability limits are reviewed. The advantages and disadvantages as well as the applicability of various prediction models for each property concerned are analyzed. Furthermore，the comparison of the neural network methods with the linear regression models，as well as the development of the neural network technique are studied. Meanwhile，the present situation and development trend of the study on design of training set，selection of molecule descriptors and validation of models are discussed. Further study of QSPR in the field of safety science is also proposed.

摘要： 介绍了神经网络及定量结构－性质相关性（QSPR）研究基本原理，综述了两者相结合在闪点、自燃点、爆炸极限等化学物质燃烧特性预测中的应用和进展。分别对各性质不同预测模型的优缺点及适用范围进行了评述。在此基础上对神经网络与线性回归方法的比较、神经网络技术的发展等进行了探讨，对实验样本设计、分子描述符选择及模型验证等的研究现状和发展趋势进行了讨论。展望了QSPR在安全研究领域的应用前景。