From 8519274487166d75b3b9ae28e61f7a7be9f7e83c Mon Sep 17 00:00:00 2001
From: Christoph Helma <helma@in-silico.ch>
Date: Thu, 27 Oct 2016 11:58:07 +0200
Subject: probability plot for classification validations

---
 test/validation-classification.rb |  2 ++
 test/validation-regression.rb     | 40 +++++++++++++++++++++++++++++++++++++--
 2 files changed, 40 insertions(+), 2 deletions(-)

(limited to 'test')

diff --git a/test/validation-classification.rb b/test/validation-classification.rb
index b71e427..c93e71f 100644
--- a/test/validation-classification.rb
+++ b/test/validation-classification.rb
@@ -11,6 +11,8 @@ class ValidationClassificationTest < MiniTest::Test
     cv = ClassificationCrossValidation.create model
     assert cv.accuracy > 0.7, "Accuracy (#{cv.accuracy}) should be larger than 0.7, this may occur due to an unfavorable training/test set split"
     assert cv.weighted_accuracy > cv.accuracy, "Weighted accuracy (#{cv.weighted_accuracy}) should be larger than accuracy (#{cv.accuracy})."
+    #p cv
+    #File.open("tmp.pdf","w+"){|f| f.puts cv.probability_plot}
   end
 
   # parameters
diff --git a/test/validation-regression.rb b/test/validation-regression.rb
index efce849..a0895f9 100644
--- a/test/validation-regression.rb
+++ b/test/validation-regression.rb
@@ -9,8 +9,9 @@ class ValidationRegressionTest < MiniTest::Test
     dataset = Dataset.from_csv_file "#{DATA_DIR}/EPAFHM.medi_log10.csv"
     model = Model::Lazar.create training_dataset: dataset
     cv = RegressionCrossValidation.create model
-    assert cv.rmse < 1.5, "RMSE #{cv.rmse} should be smaller than 1.5, this may occur due to an unfavorable training/test set split"
-    assert cv.mae < 1, "MAE #{cv.mae} should be smaller than 1, this may occur due to an unfavorable training/test set split"
+    assert cv.rmse < 1.5, "RMSE #{cv.rmse} should be smaller than 1.5, this may occur due to unfavorable training/test set splits"
+    assert cv.mae < 1.1, "MAE #{cv.mae} should be smaller than 1.1, this may occur due to unfavorable training/test set splits"
+    assert cv.percent_within_prediction_interval > 80, "Only #{cv.percent_within_prediction_interval.round(2)}% of measurement within prediction interval. This may occur due to unfavorable training/test set splits"
   end
 
   # parameters
@@ -54,4 +55,39 @@ class ValidationRegressionTest < MiniTest::Test
     assert loo.r_squared > 0.34, "R^2 (#{loo.r_squared}) should be larger than 0.034"
   end
 
+  def test_regression_loo_validation_with_feature_selection
+    dataset = OpenTox::Dataset.from_csv_file File.join(DATA_DIR,"EPAFHM.medi_log10.csv")
+    algorithms = {
+      :descriptors => {
+        :method => "calculate_properties",
+        :features => PhysChem.openbabel_descriptors,
+      },
+      :similarity => {
+        :method => "Algorithm::Similarity.weighted_cosine",
+        :min => 0.5
+      },
+      :feature_selection => {
+        :method => "Algorithm::FeatureSelection.correlation_filter",
+      },
+    }
+    model = Model::Lazar.create training_dataset: dataset, algorithms: algorithms
+    assert_raises OpenTox::BadRequestError do
+      loo = RegressionLeaveOneOut.create model
+    end
+  end
+
+  # repeated CV
+
+  def test_repeated_crossvalidation
+    dataset = OpenTox::Dataset.from_csv_file File.join(DATA_DIR,"EPAFHM.medi_log10.csv")
+    model = Model::Lazar.create training_dataset: dataset
+    repeated_cv = RepeatedCrossValidation.create model
+    repeated_cv.crossvalidations.each do |cv|
+      #assert cv.r_squared > 0.34, "R^2 (#{cv.r_squared}) should be larger than 0.034"
+      #assert_operator cv.accuracy, :>, 0.7, "model accuracy < 0.7, this may happen by chance due to an unfavorable training/test set split"
+    end
+    p repeated_cv
+    File.open("tmp.png","w+"){|f| f.puts repeated_cv.correlation_plot}
+  end
+
 end
-- 
cgit v1.2.3