遗传算法和梯度法联合训练的人工神经网络预测有机物熔点

Prediction of Melting Points of Organic Compounds Using Artificial Neural Network Trained with the Combination of Genetic Algorithm and Gradient Method

应用遗传算法和梯度法联合训练人工神经网络，使之从有机物的分子量、临界密度、正常沸点和偶极矩预测其熔点。平均误差5．1％，效果优于Joback方程和许氏方程。

The artificial neural network was trained with the combination of genetic algorithm andgradient method and used to predict the melting point based on the molecular weight, critical density,normal boiling point and dipole moment of organic compounds. The trained artificial neural networkcould make predictions of the melting points within an average error of 5.1%.