Table 2. Important Parameters Used To Establish the Models.
| models | descriptors | methodsa | parameters |
|---|---|---|---|
| model 1A | MACCS fingerprint | k-NN | the number of the closet neighbors k = 3. |
| model 1B | PubChem fingerprint | k-NN | the number of the closet neighbors k = 3 |
| model 1C | CORINA Symphony | k-NN | the number of the closet neighbors k = 10 |
| model 2A | MACCS fingerprint | DT | the maximum depth of the tree d = 0; the minimum number of samples s = 2; the number of features to consider when looking for the best split n = 13 |
| model 2B | PubChem fingerprint | DT | the maximum depth of the tree d = 0; the minimum number of samples s = 4; the number of features to consider when looking for the best split n = 30 |
| model 2C | CORINA Symphony | DT | the maximum depth of the tree d = 8; the minimum number of samples s = 2; the number of features to consider when looking for the best split n = 7 |
| model 3A | MACCS fingerprint | RF | the number of trees n = 100; the max depth of the tree d = 10; the number of features to consider when looking for the best split n = 13 |
| model 3B | PubChem fingerprint | RF | the number of trees n = 70; the max depth of the tree d = 10; the number of features to consider when looking for the best split n = 30 |
| model 3C | CORINA Symphony | RF | the number of trees n = 110; the max depth of the tree d = 6; the number of features to consider when looking for the best split n = 7 |
| model 4A | MACCS fingerprint | SVM | the optimum loss parameter C = 4.0; the kernel function parameter γ = 0.0625 |
| model 4B | PubChem fingerprint | SVM | the optimum loss parameter C = 4.0; the kernel function parameter γ = 0.0015625 |
| model 4C | CORINA Symphony | SVM | the optimum loss parameter C = 4.0; the kernel function parameter γ = 2.0 |
| model 5A | MACCS fingerprint | DNN | the number of hidden layer = 1; the number of neuron = 16 |
| model 5B | PubChem fingerprint | DNN | the number of hidden layer = 2; the number of neuron in first hidden layer = 16, the number of neuron in second hidden layer = 16 |
| model 5C | CORINA Symphony | DNN | the number of hidden layer = 1; the number of neuron = 64 |
a
NB: naive bayes; k-NN: k-nearest neighbors; DT: decision tree; RF: random forest; SVM: support vector machine; DNN: deep neural net.