运用模糊神经网络表达和预测链烷烃pVT性质

采用一种基于遗传算法的新型模糊神经网络方法研究链烷烃类化合物的pVT性质。该方法综合神经网络、遗传算法与模糊系统三种柔性智能计算技术的优点，具有良好的学习能力，不易陷入局部最小区域，学习速度较快，网络知识以模糊语言变量的形式加以表达，易于理解。用分子连接性指数对24种链烷烃化合物结构和pVT数据进行学习，进而预测另外14种未知化合物的pVT性质，较好地揭示出化合物分子结构与pVT性质之间的关系，并给出了良好的关联与预测结果。

Expression and prediction of the pVT properties of linear alkanes using fuzzy neural networks

In this paper, a new fuzzy neural network (FNN) based on genetic algorithms is proposed for studying the pVT properties of linear alkanes. The method based on fuzzy logic (FL) , neural network (NN) and genetic algorithm (GA) allows supervised learning of fuzzy rules from significant examples and is affected unsusceptibly by the problem of local extremes. The network' s knowledge base has a linguistic representation which makes it easy to understand and interpret. Using this new method and molecular connectivity index, 24 compounds are treated as a training set to extract the fuzzy knowledge base. The knowledge base extracted from examples clearly shows the relationship between the structure of compounds and their physicochemical properties. According to the training results of FNN, the pVT data of other 14 compounds are predicted. The calculated results are satisfactory. The FNN with the molecular connectivity index is a convenient and effective method to calculate the pVT data.