// This example script trains a sklearn-based model on the 
//  formation energies of a small set of compounds

// Load in a dataset of compounds
data = new data.materials.CompositionDataset
data import ./datasets/small_set.txt

// Get only the ground-state instances of each compound
data duplicates RankingDuplicateResolver minmize PropertyRanker energy_pa SimpleEntryRanker

// Define where to find elemental property data
data attributes properties directory ./lookup-data/
// Select which set of elemental properties to use for attributes
data attributes properties add set general

// Generate new attributes
data attributes generate

// Set formation energy as property to be modeled
data target delta_e

// Use the ScikitLearnRegression wrapper to train and evaluate a random forest model
//   This model was created by creating this classifier in Python and saving
//   it to a pickle file using the following script:
//     from sklearn.ensemble import RandomForestClassifier
//     clf = RandomForestClassifier(n_estimators=10, max_depth=None)
//     from sklearn.externals import joblib
//     joblib.dump(clf, 'sklearn-randomforest.pkl')
model = new models.regression.ScikitLearnRegression ./test-files/sklearn-randomforest.pkl 0
// model = new models.regression.WekaRegression functions.LinearRegression
model train $data
model crossvalidate $data 10

// Print out training and validation statistics
print model training stats
print model validation stats

exit