Merge pull request #157 from ADA110/master

Analytics Module Tests

automatminer/analytics/tests/test_core.py

@@ -0,0 +1,44 @@
+import unittest
+import os.path
+import warnings
+
+from automatminer.analytics.core import Analytics
+from automatminer.presets import get_preset_config
+from automatminer.pipeline import MatPipe
+from matminer.datasets import load_dataset
+try:
+    #from skater.core.explanations import Interpretation
+    #from skater.model import InMemoryModel
+    warnings.warn("")
+except ImportError:
+    warnings.warn("skater package not found. Please install skater to use the "
+                   "Analytics modeule/")
+    Interpretation = None
+    InMemoryModel = None
+
+
+@unittest.skipIf("CI" in os.environ.keys(), "Test too intensive for CircleCI.")
+class TestAnalytics(unittest.TestCase):
+
+    def setUp(self):
+        df = load_dataset('elastic_tensor_2015')
+        df = df[["formula", "K_VRH"]]
+        df = df.rename({"formula": "composition"}, axis=1)
+        self.config = get_preset_config("debug")
+        fitted_pipeline = MatPipe(**self.config).fit(df, "K_VRH")
+
+        self.analytics = Analytics(predictive_model=fitted_pipeline)
+
+
+    def test_get_feature_importance(self):
+        feature_importance = self.analytics.get_feature_importance()
+        print(feature_importance)
+        self.assertTrue(not feature_importance.empty)
+
+
+    def test_plot_partial_dependence(self):
+        feature_importance = self.analytics.get_feature_importance()
+        for feature in feature_importance.index[len(feature_importance.index)-15:]:
+            self.analytics.plot_partial_dependence(feature, save_plot=True, show_plot=False)
+        self.assertTrue(os.path.isfile("t_pdp.png"))
+        self.assertTrue(os.path.isfile("m_pdp.png"))