Integrated System of Nonlinear Modelling to Predict Normal Boiling Point of Cyclic Hydrocarbons Using Inferential Statistical

Abstract

An integrated system of predictive nonlinear contribution models (R²_predictive » 0.99) using inferential statistical methods is shown in this work with the intention of inferring the boiling point in cyclic hydrocarbons that could be determined from new molecular structures by molecular reconstruction methods and/or pseudo-compounds, as a cheaper and faster tool in industrial applications, environmental studies and health risk assessments. This model relates geometric molecular descriptors with one intensive physical property for a training set of 100 molecules. Additionally, this work shows the reduction of database dimensional space highly correlated up to 99.2% using statistical methods obtaining information relevant to predictive contribution models.