diff --git a/tests/test_timeseriesnormalizer.py b/tests/test_timeseriesnormalizer.py
new file mode 100644
index 0000000..cd11fa0
--- /dev/null
+++ b/tests/test_timeseriesnormalizer.py
@@ -0,0 +1,47 @@
+import numpy as np
+import pandas as pd
+import pytest
+
+from apyxl import TimeSeriesNormalizer
+
+
+@pytest.fixture
+def sample_data():
+    n = 8760
+    time = pd.date_range(start='2024-01-01', freq='h', periods=n)
+
+    cov = [[1, 0.7], [0.7, 1]]
+    mean = [0, 5]
+
+    df = np.random.multivariate_normal(cov=cov, mean=mean, size=n)
+    df[:, 1] *= 2
+
+    df[6000:7000, 1] += 2
+
+    return pd.DataFrame(df, columns=['a', 'b'], index=time)
+
+
+def test_time_series_normalizer_normalize(sample_data):
+    tsn = TimeSeriesNormalizer(freq_trend='1d')
+    trend = tsn.normalize(sample_data, target='b')
+
+    assert isinstance(trend, pd.Series)
+    assert len(trend) == len(sample_data)
+    assert trend.index.equals(sample_data.index)
+
+
+def test_time_series_normalizer_output_shape(sample_data):
+    tsn = TimeSeriesNormalizer(freq_trend='1d')
+    trend = tsn.normalize(sample_data, target='b')
+
+    assert trend.shape == (8760,)
+
+
+def test_time_series_normalizer_with_different_freq(sample_data):
+    tsn_hourly = TimeSeriesNormalizer(freq_trend='1h')
+    tsn_daily = TimeSeriesNormalizer(freq_trend='1d')
+
+    trend_hourly = tsn_hourly.normalize(sample_data, target='b')
+    trend_daily = tsn_daily.normalize(sample_data, target='b')
+
+    assert not np.allclose(trend_hourly, trend_daily)  # The trends should be different
diff --git a/tests/test_xgbwrapper.py b/tests/test_xgbwrapper.py
index 007798d..8a70ad7 100644
--- a/tests/test_xgbwrapper.py
+++ b/tests/test_xgbwrapper.py
@@ -4,7 +4,8 @@
 import shap
 from sklearn.datasets import load_diabetes
 
-from apyxl import MissingInputError, NotFittedError, XGBRegressorWrapper
+from apyxl import XGBRegressorWrapper
+from apyxl._misc import MissingInputError, NotFittedError
 
 
 @pytest.fixture