Thumb Rules for Machine Learning-Based Modelling

The parity plot of the desired and predicted values of y is shown in Fig. 3. 64. The value of 0.96 for the coefficient of correlation between the desired and model-predicted values of y (training and test sets) indicates good prediction accuracy of the model. 

8 Thumb Rules for Machine Learning-Based Modelling 

Use as much example data as possible for training the ML model since a model trained on a large data set is likely to possess better prediction and generalization ability. Note that data adequacy depends on various factors including the dimensionality of the system being modelled. 

ANNs and SVR can handle qualitative inputs and/or outputs provided these are appropriately represented in numeric quantities. 

3 Machine Learning Methods in Chemoinformatics for Drug Discovery 

---