// This example script ranks attributes based on their ability 
//  to predict formation energy for a set of compounds, and then generates
//  a reduced subset of attributes with Weka

// Load in a dataset of compound properties
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

// Print out 10 attributes with the best correlation coefficient
data attributes rank 10 SpearmansCorrelationEvaluator

// Generate a reduced list of attributes with an AttributeSelector
selector = new attributes.selectors.WekaAttributeSelector -eval CfsSubsetEval -M -search BestFirst -D 1
selector train $data
selector run $data

exit