The Prediction of Boiling Point and Melting Point of Aniline Derivatives by Multiple- Layer Perceptron and Non-Linear Partial Least Square Methods

Quantitative Structure-Properties Relationship (QSPR) was used to study melting and boiling points of aniline derivatives. In the present study, using QSPR methodology, some aniline derivatives were analyzed and the data were divided into two categories of train and test (validation) sets using random selection methods. To find the relation between descriptors and mentioned properties, Multiple-Layer Perceptron (MLP) andNonlinearpartial least squares (nPLS) were used. Different QSAR models show that the MLP approach for prediction of boiling point (bp)with N=75, R= 0.417,R2 adj= 0.173 and RMSE=0.52 give the best outcome, and also QSPR methodology show that MLP approach with N=62, R= 0.796, R2 adj= 0.633 and RMSE=0.39 is in good agreement with the experimental melting point (mp).The high correlation between experimental and predicted mpand bp values of aniline derivatives was observed, indicating the validation and the good quality of the resulting MLP approach.