Abstarct: In the first section, quantitative structure-activity relationship (QSAR) models of 56 compounds of CDK2 inhibitors were constructed using linear (multiple linear regression; MLR) and nonlinear (artificial neural network; ANN) methods. The cyclin dependent kinases (CDKs) are compounds which play a fundamental role in cell cycle regulation and inhibit the overactive CDK2 which result in excessive reproduction of cells and finally in cancer. The prediction ability of the models was evaluated using the validation and test data sets, leave-one-out cross validation method and Y-randomization. The mean squared error (MSE) for test set by ANN and MLR models are 0.063 and 0.069, respectively. In the second section, quantitative structure-property relationship (QSPR) models of retention indices of 196 monomethyl alkanes were constructed using two linear (multiple linear regression; MLR) and nonlinear (artificial neural network; ANN) methods. The prediction ability of the models was evaluated using the validation and test data sets, leave-one-out cross validation method and Y-randomization. The mean squared error (MSE) for test set by ANN and MLR models are 4.21 and 28.96, respectively. The results of both sections show better prediction ability for nonlinear models.