Skip to main content

Table 5 Summary of the classifier models that have been used in the two phases of the virtual screening experiment

From: Spectrophores as one-dimensional descriptors calculated from three-dimensional atomic properties: applications ranging from scaffold hopping to multi-target virtual screening

Phase Classifier with parametersa Precision ± SDb
Phase 1 (binary classification) Soft voting classifier with 2 underlying models:
Random forest classifier:
 criterion = ‘entropy’; max_features = ‘log2’; n_estimators = 30 
k-nearest neighbors classifier:
 n_neighbors = 28; weights = ‘uniform’
0.80 ± 0.07
Phase 2 (multiclass) Extra Trees classifier:
 max_features = None; criterion = ‘gini’; n_estimators = 90; min_samples_leaf = 1
0.63 ± 0.01
  1. aParameters as implemented in the scikit-learn package
  2. bMean and standard deviation calculated from tenfold cross-validation