Multiple Linear Regressionamp;horbar;Artificial Neural Network Hybrid Model Predicting Liquid Density of Pure Organic Compound for Normal Boiling Point

摘要

The present invention is hydrogen (H), carbon (C), nitrogen (N), oxygen (O) , sulfur (S), such as five kinds of elements within the configuration is other than hydrogen atom is 25 or less of the number of molecules with a pure organic compound consisting of a normal boiling point in the density of the liquid (liquid density at normal boiling point) higher accuracy in prediction that provides a mathematical model. The model is for a plurality of organic compound satisfying the above conditions are experimental values of the liquid density is known, various presenter molecule (molecular descriptor) some of the independent variables, dependent variables, the liquid density at the normal boiling point Many multi-linear regression model (multiple linear regression model) obtained after using the genetic algorithm (genetic algorithm) that one of the best, receives the value of the molecules contained in the presenter, the model of the normal boiling point of a liquid density the output of ANN (artificial neural network) to further improve the prediction performance was composed by multiple linear regression-ANN hybrid model (hybrid model) as QSPR (quantitative structure-property relationship) model and an example, the model to include molecules If known, the specific value of the presenter that some molecules way, gives to predict the density of the liquid in a normal boiling point of the pure compound made by this molecule. As such, the present invention can maintain the cost and time savings by giving the experiment to provide a way to predict the value of liquid density in the normal boiling point reliable for the number of the organic compound of the experimental conditions is unknown, the related industry lays effects such as the ease of research and development activities. ;