def test_exog_seesaw(self):
exog = Series.from_array([
[0, 1],
[1, -1],
[2, 1],
[3, -1],
[4, 1],
[5, -1],
[6, 1],
[7, -1],
[8, 1],
[9, -1],
])
endog = TestSeriesBuilder.linear(10, 0, 1).build()
factory = NodeFactory.transformer('test', 'SARIMAX')
factory.set_param_value('p', '1')
factory.set_param_value('d', 1)
factory.set_param_value('output', 'resid')
sarimax = factory.build()
result, debug_info = sarimax.transform([endog, exog], True)
expected_series = [0] * 8
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertAlmostEqual(expected_series[i], actual_series[i], 2)
# With debug info
self.assertEqual(set(['summary', 'offset_start', 'exog_coeff_x1']),
set(debug_info.keys()))
def test_linear_series_lags(self):
series = TestSeriesBuilder.linear(10, 0, 1).build()
factory = NodeFactory.transformer('test', 'Identity')
factory.set_param_value('mean', True)
factory.set_param_value('stddev', True)
factory.set_param_value('adf_test', True)
factory.set_param_value('normality_test', True)
factory.set_param_value('histogram', True)
factory.set_param_value('histogram_bins', 10)
factory.set_param_value('cum_histogram', True)
factory.set_param_value('cum_histogram_bins', 10)
factory.set_param_value('acf', True)
factory.set_param_value('acf_lags', 10)
factory.set_param_value('pacf', True)
factory.set_param_value('pacf_lags', 10)
identity = factory.build()
result, debug_info = identity.transform([series], True)
expected_series = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertEqual(expected_series[i], actual_series[i])
# With debug info
self.assertEqual(EXPECTED_KEYS, set(debug_info.keys()))
self.assertEqual(4.5, debug_info['Mean'])
self.assertAlmostEqual(3.02, debug_info['Std. dev.'], 1)
# No debug info
result, debug_info = identity.transform([series], False)
self.assertEqual([], list(debug_info.keys()))
def test_base_derivative_non_linear(self):
self.maxDiff = None
series1 = Series.from_array([
[0, 1],
[1, 2],
[2, 4],
[3, 0],
[4, 16],
])
factory = NodeFactory.transformer('test', 'Derivative')
factory.set_param_value('lag', '1')
factory.set_param_value('metric', 'sub')
diff = factory.build()
result, debug_info = diff.transform([series1], False)
expected_series = Series.from_array([
[1, 1],
[2, 2],
[3, -4],
[4, 16],
])
self.assertEqual(list(expected_series.pdseries.index),
list(result.index))
self.assertEqual(list(expected_series.pdseries.values),
list(result.values))
def test_rolling_aggregate_mean(self):
factory = NodeFactory.transformer('test', 'RollingAggregate')
factory.set_param_value('window', 2)
factory.set_param_value('center', False)
factory.set_param_value('min_periods', 0)
factory.set_param_value('agg_method', 'mean')
factory.add_source(InputRef('input'))
ram = factory.build()
self.case(ram, [0.0, 0.5, 1.5, 2.5, 3.5, 4.5, 5.5, 6.5, 7.5, 8.5, 9.0, 8.5, 7.5, 6.5, 5.5, 4.5, 3.5, 2.5, 1.5, 0.5])
def test_divide_discard_fill_zero_div(self):
series1 = Series.from_array([[1, 1], [2, 2], [3, 3]])
series2 = Series.from_array([[1, 2], [2, 0], [3, 5]])
factory = NodeFactory.transformer('test', 'Divide')
factory.set_param_value('zero_div', 0)
divide = factory.build()
result, debug_info = divide.transform([series1, series2], False)
expected_series = pd.Series([1 / 2, 0.0, 3 / 5],
index=series1.pdseries.index)
self.assertEqual(list(expected_series.index), list(result.index))
self.assertEqual(list(expected_series.values), list(result.values))
def test_exponential_moving_average_mean(self):
factory = NodeFactory.transformer('test', 'ExponentialMovingAverage')
factory.set_param_value('span', '4')
factory.set_param_value('min_periods', '')
factory.set_param_value('recursive', False)
factory.set_param_value('agg_method', 'mean')
factory.add_source(InputRef('input'))
ewma = factory.build()
expected_debug_info = {'alpha': 0.4}
self.case(ewma, [2.095, 2.487, 2.283, 1.755, 1.041],
expected_debug_info)
def test_dropout_mean_prop(self):
factory = NodeFactory.transformer('test', 'Dropout')
factory.set_param_value('context_window', 2)
factory.set_param_value('dropout_window', 1)
factory.set_param_value('center', False)
factory.set_param_value('min_periods', None)
factory.set_param_value('agg_method', 'mean')
factory.set_param_value('combine_method', 'prop')
factory.set_param_value('combine_method_order', 'dropout-context')
factory.add_source(InputRef('input'))
ram = factory.build()
self.case(ram, [1, 1, 3, 1, 1 / 3, 1, 1])
def test_multi_rolling_aggregate_correlation(self):
factory = NodeFactory.transformer('test', 'MultiRollingAggregate')
factory.set_param_value('window', '2')
factory.set_param_value('center', False)
factory.set_param_value('min_periods', '0')
factory.set_param_value('agg_method', 'correlation_pearson')
factory.add_source(InputRef('lhs'))
factory.add_source(InputRef('rhs'))
ram = factory.build()
s1 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 4], [4, 5]])
s2 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 2], [4, 1]])
self.case(ram, s1, s2, list(s1.pdseries.index)[1:], [1, 1, -1, -1])
def test_dropout_centered_mean_sub(self):
factory = NodeFactory.transformer('test', 'Dropout')
factory.set_param_value('context_window', 4)
factory.set_param_value('dropout_window', 2)
factory.set_param_value('center', True)
factory.set_param_value('min_periods', None)
factory.set_param_value('agg_method', 'mean')
factory.set_param_value('combine_method', 'sub')
factory.set_param_value('combine_method_order', 'context-dropout')
factory.add_source(InputRef('input'))
ram = factory.build()
self.case(ram, [1, 0, -2, 0, 1])
def test_multi_rolling_aggregate_mismatching_start(self):
factory = NodeFactory.transformer('test', 'MultiRollingAggregate')
factory.set_param_value('window', '1')
factory.set_param_value('center', False)
factory.set_param_value('min_periods', '0')
factory.set_param_value('agg_method', 'proportion')
factory.add_source(InputRef('lhs'))
factory.add_source(InputRef('rhs'))
ram = factory.build()
s1 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 4]])
s2 = Series.from_array([[1, 5], [2, 5], [3, 5]])
self.case(ram, s1, s2, list(s2.pdseries.index), [2 / 5, 3 / 5, 4 / 5])
def test_multi_rolling_aggregate_proportion_window_min_periods_default(
self):
factory = NodeFactory.transformer('test', 'MultiRollingAggregate')
factory.set_param_value('window', '2')
factory.set_param_value('center', False)
factory.set_param_value('min_periods', '')
factory.set_param_value('agg_method', 'proportion')
factory.add_source(InputRef('lhs'))
factory.add_source(InputRef('rhs'))
ram = factory.build()
s1 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 4]])
s2 = Series.from_array([[0, 5], [1, 5], [2, 5], [3, 5]])
self.case(ram, s1, s2, list(s1.pdseries.index[1:]), [0.3, 0.5, 0.7])
def test_seesaw(self):
series = Series.from_array([[0, 1], [1, -1], [2, 1], [3, -1], [4, 1],
[5, -1]])
factory = NodeFactory.transformer('test', 'SARIMAX')
factory.set_param_value('p', '1')
factory.set_param_value('output', 'resid')
sarimax = factory.build()
result, debug_info = sarimax.transform([series], True)
expected_series = [0] * 5
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertAlmostEqual(expected_series[i], actual_series[i], 2)
def test_analysis_without_debug(self):
self.maxDiff = None
series = self.build_triangle()
factory = NodeFactory.transformer('test_node', 'RollingAggregate')
factory.set_param_value('window', 5)
factory.set_param_value('center', False)
factory.set_param_value('min_periods', 0)
factory.set_param_value('agg_method', 'max')
factory.add_source(InputRef('input'))
node = factory.build()
pipeline = Pipeline([node])
analyzer = Analyzer(pipeline=pipeline, debug=False)
analysis = analyzer.analyze({'input':series})
expected = {
"anomalies": [],
"series": {
"input": [
[0, 0.0],
[1000, 1.0],
[2000, 2.0],
[3000, 3.0],
[4000, 4.0],
[5000, 5.0],
[6000, 6.0],
[7000, 7.0],
[8000, 8.0],
[9000, 9.0],
[10000, 9.0],
[11000, 8.0],
[12000, 7.0],
[13000, 6.0],
[14000, 5.0],
[15000, 4.0],
[16000, 3.0],
[17000, 2.0],
[18000, 1.0],
[19000, 0.0],
]
},
}
self.assertEqual(expected, analysis.output_format())
def test_std_normalize(self):
series = self.build_triangle()
self.assertEqual(
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0],
series.as_list())
factory = NodeFactory.transformer('std_normalize', 'StdNormalize')
factory.add_source(InputRef('input'))
stdnormalize = factory.build()
self.case(series, stdnormalize, [
-1.5270292013639366, -1.1876893788386174, -0.8483495563132981,
-0.5090097337879789, -0.16966991126265962, 0.16966991126265962,
0.5090097337879789, 0.8483495563132981, 1.1876893788386174,
1.5270292013639366, 1.5270292013639366, 1.1876893788386174,
0.8483495563132981, 0.5090097337879789, 0.16966991126265962,
-0.16966991126265962, -0.5090097337879789, -0.8483495563132981,
-1.1876893788386174, -1.5270292013639366
])
def test_base_derivative_lag2(self):
series1 = TestSeriesBuilder.linear(5, 0, 1).build()
factory = NodeFactory.transformer('test', 'Derivative')
factory.set_param_value('lag', '2')
factory.set_param_value('metric', 'sub')
diff = factory.build()
result, debug_info = diff.transform([series1], False)
expected_series = Series.from_array([
[2, 2],
[3, 2],
[4, 2],
])
self.assertEqual(list(expected_series.pdseries.index),
list(result.index))
self.assertEqual(list(expected_series.pdseries.values),
list(result.values))
def test_boxcox_fixed_lambda(self):
series = self.build_triangle()
factory = NodeFactory.transformer('boxcox', 'BoxCox')
factory.set_param_value('lambda', 1)
factory.add_source(InputRef('input'))
boxcox = factory.build()
expected_debug = {
'lambda': 1,
'fitted_lambda': 1,
}
self.maxDiff = None
self.case(series, boxcox, [
0.0, 5.0, 10.000000000000002, 14.999999999999998, 20.0, 20.0,
18.999999999999996, 17.999999999999996, 16.999999999999996, 16.0,
14.999999999999998, 14.0, 12.999999999999996, 12.0, 11.0,
10.000000000000002, 9.000000000000002, 8.000000000000002,
6.999999999999998, 5.999999999999999, 5.0, 3.999999999999999, 3.0,
2.0000000000000004, 1.0
], expected_debug)
def test_linear_series_lags_fitted(self):
series = TestSeriesBuilder.linear(10, 0, 1).build()
factory = NodeFactory.transformer('test', 'AutoRegression')
factory.set_param_value('lags', '2')
factory.set_param_value('period', '')
factory.set_param_value('output', 'predicted')
ar = factory.build()
result, debug_info = ar.transform([series], True)
# print(debug_info)
expected_series = series.as_list()[2:]
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertAlmostEqual(expected_series[i], actual_series[i], 2)
# With debug info
self.assertEqual(set(['summary']), set(debug_info.keys()))
# No debug info
result, debug_info = ar.transform([series], False)
self.assertEqual([], list(debug_info.keys()))
def test_fft_filter(self):
series = self.build_triangle()
self.assertEqual(
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0],
series.as_list())
factory = NodeFactory.transformer('fft_filter', 'FFTFilter')
factory.set_param_value('cutoff', '50%')
factory.set_param_value('output', 'resid')
factory.add_source(InputRef('input'))
fft_filter = factory.build()
self.case(series, fft_filter, [
0.002508563093691407, 0.9927198663527221, 2.0113390739957944,
2.9857119332223716, 4.015838444032453, 4.984161555967546,
6.014288066777628, 6.988660926004205, 8.00728013364728,
8.997491436906309, 8.99749143690631, 8.007280133647278,
6.988660926004206, 6.014288066777628, 4.984161555967547,
4.015838444032454, 2.985711933222371, 2.0113390739957935,
0.9927198663527218, 0.002508563093691407
])
def test_boxcox_implied_lambda(self):
series = self.build_triangle()
factory = NodeFactory.transformer('boxcox', 'BoxCox')
# factory.set_param_value('lambda', 1)
factory.add_source(InputRef('input'))
boxcox = factory.build()
expected_debug = {
'lambda': None,
'fitted_lambda': 0.7569500835813396,
}
self.case(series, boxcox, [
0.0, 3.8069844496343204, 6.792535931588043, 9.45326887138231,
11.915823446529231, 11.915823446529231, 11.435878585907972,
10.950062851581707, 10.457989331193566, 9.959222501860888,
9.45326887138231, 8.939565111800931, 8.417462798503934,
7.88620846728669, 7.344917074353449, 6.792535931588043,
6.227794497220499, 5.649132457366781, 5.054593159635213,
4.441659075192572, 3.8069844496343204, 3.1459316503179107,
2.45169446968289, 1.7134270251968304, 0.9114396216016829
], expected_debug)
def test_linear_series_lags(self):
series = TestSeriesBuilder.linear(10, 0, 1).build()
factory = NodeFactory.transformer('test', 'SARIMAX')
factory.set_param_value('p', '1')
factory.set_param_value('d', 1)
# factory.set_param_value('q', '3')
factory.set_param_value('output', 'resid')
sarimax = factory.build()
result, debug_info = sarimax.transform([series], True)
expected_series = [0] * 8
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertAlmostEqual(expected_series[i], actual_series[i], 2)
# With debug info
self.assertEqual(set(['summary', 'offset_start']),
set(debug_info.keys()))
# No debug info
result, debug_info = sarimax.transform([series], False)
self.assertEqual([], list(debug_info.keys()))
def test_analysis_with_debug(self):
self.maxDiff = None
series = self.build_triangle()
factory = NodeFactory.transformer('test_node', 'RollingAggregate')
factory.set_param_value('window', 5)
factory.set_param_value('center', False)
factory.set_param_value('min_periods', 0)
factory.set_param_value('agg_method', 'max')
factory.add_source(InputRef('input'))
node = factory.build()
pipeline = Pipeline([node])
analyzer = Analyzer(pipeline=pipeline, debug=True)
analysis = analyzer.analyze({'input': series})
actual_output = analysis.output_format()
expected_file = os.path.join(os.path.dirname(__file__), 'resources/analysis/expected_simplified.json')
# Uncomment to fix test
# print(json.dumps(actual_output, indent=2), file=open(expected_file, 'w'))
expected_output = json.loads(Path(expected_file).read_text())
self.assertEqual(expected_output, actual_output)