// Cluster a set of compounds according to a subset of their attributes 

// Load in a dataset of compound properties
data = new data.materials.CompositionDataset
data import ./datasets/small_set.txt
data filter exclude ContainsElementFilter He Ne Ar Kr Xe
data target delta_e
data attributes properties directory ./lookup-data/
data attributes properties add set general
data attributes generate

// Create an attribute selector
selector = new attributes.selectors.WekaAttributeSelector -eval CfsSubsetEval -M -search BestFirst -D 1

// Create a simple clusterer using Weka
clusterer = new cluster.WekaClusterer SimpleKMeans -N 2 -I 500 
clusterer selector $selector

// Train this cluster on the full dataset
clusterer train $data

// Using this clusterer, extract all entries from data
//  that fall in cluster #0, save as cluster 0
cluster0 = clusterer get 0 $data 
print cluster0 details

// Print out information about centroids of clusters
print clusterer stats clust

// Split "data" into clusters, then save each with a
//  filename of clusters<#> in "stats" format
clusterer split $data cluster stats


exit