def test_metrics(testdata_df):
analyzer = analyze.PredictionAnalyzer(testdata_df)
assert analyzer.y_true == testdata_df["target"].tolist()
assert analyzer.y_pred == testdata_df["max_likelihood"].tolist()
assert isinstance(analyzer.mean_loss, float)
assert np.sum(analyzer.support) == len(testdata_df)
assert len(analyzer.tps) == len(testdata_df)
def test_save_load_analyzer(workspace, testdata_df):
test_set = "rwc"
analyzer = analyze.PredictionAnalyzer(testdata_df, test_set=test_set)
save_path = os.path.join(workspace, "analyzer.pkl")
analyzer.save(save_path)
assert os.path.exists(save_path)
# Prove you can load it back in, too.
analyzer2 = analyze.PredictionAnalyzer.load(save_path, test_set)
assert analyzer.y_true == analyzer2.y_true
assert analyzer.y_pred == analyzer2.y_pred
def test_dataset_summary(testdata_df):
analyzer = analyze.PredictionAnalyzer(testdata_df)
scores = analyzer.dataset_summary()
assert set(scores.index) == \
set(["overall", "rwc", "uiowa", "philharmonia"])
assert scores.shape[1] == 4
def test_dataset_class_wise(testdata_df):
analyzer = analyze.PredictionAnalyzer(testdata_df)
scores = analyzer.dataset_class_wise()
assert set(scores.index.levels[0]) == \
set(["overall", "rwc", "uiowa", "philharmonia"])
assert scores.shape[1] == 4
def test_summary_scores(testdata_df):
analyzer = analyze.PredictionAnalyzer(testdata_df, test_set="rwc")
scores = analyzer.summary_scores()
assert set(scores.index) == set(["precision", "recall",
"f1score", "accuracy"])
assert all(np.isfinite(scores))
def test_class_wise_scores(testdata_df):
analyzer = analyze.PredictionAnalyzer(testdata_df, test_set="rwc")
scores = analyzer.class_wise_scores()
assert set(scores.columns) == set(["precision", "recall",
"f1score", "support"])
assert all(np.isfinite(scores))
def test_prediction_analyzer(testdata_df):
analyzer = analyze.PredictionAnalyzer(testdata_df, test_set="rwc")
assert analyzer is not None