diff options
Diffstat (limited to 'r-util.rb')
| -rw-r--r-- | r-util.rb | 52 |
1 files changed, 15 insertions, 37 deletions
@@ -146,14 +146,9 @@ class RUtilTest < Test::Unit::TestCase def stratified_split unless defined?@@strat @@split_ratio = 0.05 - @@split_has_duplicates = false #hamster has no duplicates -# res = @@rutil.stratified_split(@@hamster,0,@@split_ratio,1) -# @@resources += [ res[0].uri, res[1].uri ] -# @@strat = { :data => @@hamster, :split1 => res[0], :split2 => res[1] } - data_combined = OpenTox::Dataset.merge(@@hamster,@@hamster_features,{},@@subjectid) - res1, res2 = @@rutil.stratified_split(data_combined,{},0,@@split_ratio,@@subjectid,1) - @@resources += [ data_combined.uri, res1.uri, res2.uri ] - @@strat = {:data => data_combined, :split1 => res1, :split2 => res2 } + res1, res2 = @@rutil.stratified_split(@@hamster_features,{},0,@@split_ratio,@@subjectid,1) + @@resources += [ res1.uri, res2.uri ] + @@strat = {:data => @@hamster_features, :split1 => res1, :split2 => res2 } end @@strat end @@ -165,31 +160,22 @@ class RUtilTest < Test::Unit::TestCase size1 = split[:split1].compounds.size size2 = split[:split2].compounds.size assert_equal size,(split[:split1].compounds+split[:split2].compounds).uniq.size - unless @@split_has_duplicates - assert_equal (@@split_ratio*size).round,size1, - "Dataset #{size} should be split into #{(@@split_ratio*size).round}/#{size-(@@split_ratio*size).round}"+ - " (exact: #{@@split_ratio*size}), instead: #{size1}/#{size2}" - end + assert_equal (@@split_ratio*size).round,size1, + "Dataset #{size} should be split into #{(@@split_ratio*size).round}/#{size-(@@split_ratio*size).round}"+ + " (exact: #{@@split_ratio*size}), instead: #{size1}/#{size2}" split[:data].compounds.each do |c| include1 = split[:split1].compounds.include?(c) include2 = split[:split2].compounds.include?(c) - unless @@split_has_duplicates - assert(((include1 and !include2) or (!include1 and include2))) - else - assert((include1 or include2)) - end + assert(((include1 and !include2) or (!include1 and include2))) end end def test_k_fold_stratified_split puts "test_k_fold_stratified_split" - data_combined = OpenTox::Dataset.merge(@@hamster,@@hamster_features,{},@@subjectid) - num_duplicates = 0 #hamster has no duplicates num_folds = 10 - avg_split_size = (data_combined.compounds.size+num_duplicates)/num_folds.to_f + avg_split_size = (@@hamster_features.compounds.size)/num_folds.to_f - @@resources += [ data_combined.uri ] - train, test = @@rutil.stratified_k_fold_split(data_combined,{},0,num_folds,@@subjectid,1) + train, test = @@rutil.stratified_k_fold_split(@@hamster_features,{},0,num_folds,@@subjectid,1) @@resources += (train + test).collect{ |r| r.uri } [train, test].each do |result| assert result.is_a?(Array) @@ -204,28 +190,20 @@ class RUtilTest < Test::Unit::TestCase sum_test+=test[i].compounds.size compounds_test += test[i].compounds - assert_equal (test[i].compounds.size+train[i].compounds.size),(data_combined.compounds.size+num_duplicates) + assert_equal (test[i].compounds.size+train[i].compounds.size),(@@hamster_features.compounds.size) compounds = (test[i].compounds + train[i].compounds) - if num_duplicates==0 - assert_equal compounds.sort,data_combined.compounds.sort - else - assert_equal compounds.uniq.sort,data_combined.compounds.sort - end - end - assert_equal sum_test,(data_combined.compounds.size+num_duplicates) - if num_duplicates==0 - assert_equal compounds_test.sort,data_combined.compounds.sort - else - assert_equal compounds_test.uniq.sort,data_combined.compounds.sort + assert_equal compounds.sort,@@hamster_features.compounds.sort end + assert_equal sum_test,(@@hamster_features.compounds.size) + assert_equal compounds_test.sort,@@hamster_features.compounds.sort end def test_feature_value_plot puts "feature_value_plot" split = stratified_split data = split[:data] - dataset1 = data.split( data.compounds[0..4], data.features.keys, {}, @@subjectid) - dataset2 = data.split( data.compounds[5..-1], data.features.keys, {}, @@subjectid) + dataset1 = data.split( (0..4).to_a, data.features.keys, {}, @@subjectid) + dataset2 = data.split( (5..(data.compounds.size-1)).to_a, data.features.keys, {}, @@subjectid) @@resources += [dataset1.uri, dataset2.uri] files = [] #plot |