Standard Deviation

Just The Code

>>> import nannyml as nml
>>> from IPython.display import display

>>> reference_df, analysis_df, analysis_targets_df = nml.load_synthetic_car_loan_dataset()
>>> display(reference_df.head())

>>> feature_column_names = [
...     'car_value', 'debt_to_income_ratio', 'driver_tenure'
>>> ]
>>> calc = nml.SummaryStatsStdCalculator(
...     column_names=feature_column_names,
>>> )

>>> calc.fit(reference_df)
>>> results = calc.calculate(analysis_df)
>>> display(results.filter(period='all').to_df())

>>> for column_name in results.column_names:
...     results.filter(column_names=column_name).plot().show()

Walkthrough

The Standard Deviation value calculation is straightforward. For each chunk NannyML calculates the standard deviation for all selected numerical columns. The resulting values from the reference data chunks are used to calculate the alert thresholds. The standard deviation value results from the analysis chunks are compared against those thresholds and generate alerts if applicable.

We begin by loading the synthetic car loan dataset provided by the NannyML package.

>>> import nannyml as nml
>>> from IPython.display import display

>>> reference_df, analysis_df, analysis_targets_df = nml.load_synthetic_car_loan_dataset()
>>> display(reference_df.head())

	id	car_value	salary_range	debt_to_income_ratio	loan_length	repaid_loan_on_prev_car	size_of_downpayment	driver_tenure	repaid	timestamp	y_pred_proba	y_pred
0	0	39811	40K - 60K €	0.63295	19	False	40%	0.212653	1	2018-01-01 00:00:00.000	0.99	1
1	1	12679	40K - 60K €	0.718627	7	True	10%	4.92755	0	2018-01-01 00:08:43.152	0.07	0
2	2	19847	40K - 60K €	0.721724	17	False	0%	0.520817	1	2018-01-01 00:17:26.304	1	1
3	3	22652	20K - 20K €	0.705992	16	False	10%	0.453649	1	2018-01-01 00:26:09.456	0.98	1
4	4	21268	60K+ €	0.671888	21	True	30%	5.69526	1	2018-01-01 00:34:52.608	0.99	1

The SummaryStatsStdCalculator class implements the functionality needed for standard deviation values calculations. We need to instantiate it with appropriate parameters:

• column_names: A list with the names of columns to be evaluated.

• timestamp_column_name (Optional): The name of the column in the reference data that contains timestamps.

• chunk_size (Optional): The number of observations in each chunk of data used. Only one chunking argument needs to be provided. For more information about chunking configurations check out the chunking tutorial.

• chunk_number (Optional): The number of chunks to be created out of data provided for each period.

• chunk_period (Optional): The time period based on which we aggregate the provided data in order to create chunks.

• chunker (Optional): A NannyML Chunker object that will handle the aggregation provided data in order to create chunks.

• threshold (Optional): The threshold strategy used to calculate the alert threshold limits. For more information about thresholds, check out the thresholds tutorial.

>>> feature_column_names = [
...     'car_value', 'debt_to_income_ratio', 'driver_tenure'
>>> ]
>>> calc = nml.SummaryStatsStdCalculator(
...     column_names=feature_column_names,
>>> )

Next, the fit() method needs to be called on the reference data, which provides the baseline that the analysis data will be compared with for alert generation. Then the calculate().calculate` method will calculate the data quality results on the data provided to it.

The results can be filtered to only include a certain data period, method or column by using the filter method. You can evaluate the result data by converting the results into a DataFrame, by calling the to_df() method. By default this will return a DataFrame with a multi-level index. The first level represents the column, the second level represents resulting information such as the data quality metric values, the alert thresholds or the associated sampling error.

>>> calc.fit(reference_df)
>>> results = calc.calculate(analysis_df)
>>> display(results.filter(period='all').to_df())

	chunk key	chunk_index	start_index	end_index	start_date	end_date	period	car_value value	sampling_error	upper_confidence_boundary	lower_confidence_boundary	upper_threshold	lower_threshold	alert	debt_to_income_ratio value	sampling_error	upper_confidence_boundary	lower_confidence_boundary	upper_threshold	lower_threshold	alert	driver_tenure value	sampling_error	upper_confidence_boundary	lower_confidence_boundary	upper_threshold	lower_threshold	alert
0	[0:4999]	0	0	4999			reference	20403.1	271.992	21219.1	19587.1	20978.6	19816.9	False	0.154082	0.00124756	0.157824	0.150339	0.159073	0.151493	False	2.29725	0.0173422	2.34928	2.24523	2.3309	2.27314	False
1	[5000:9999]	1	5000	9999			reference	20527.4	271.992	21343.4	19711.5	20978.6	19816.9	False	0.157558	0.00124756	0.161301	0.153816	0.159073	0.151493	False	2.29714	0.0173422	2.34916	2.24511	2.3309	2.27314	False
2	[10000:14999]	2	10000	14999			reference	20114.8	271.992	20930.8	19298.9	20978.6	19816.9	False	0.15577	0.00124756	0.159513	0.152028	0.159073	0.151493	False	2.29814	0.0173422	2.35016	2.24611	2.3309	2.27314	False
3	[15000:19999]	3	15000	19999			reference	20434.1	271.992	21250.1	19618.1	20978.6	19816.9	False	0.156043	0.00124756	0.159786	0.152301	0.159073	0.151493	False	2.28299	0.0173422	2.33502	2.23097	2.3309	2.27314	False
4	[20000:24999]	4	20000	24999			reference	20212.7	271.992	21028.7	19396.7	20978.6	19816.9	False	0.155773	0.00124756	0.159515	0.15203	0.159073	0.151493	False	2.31656	0.0173422	2.36859	2.26453	2.3309	2.27314	False
5	[25000:29999]	5	25000	29999			reference	20714.8	271.992	21530.7	19898.8	20978.6	19816.9	False	0.156099	0.00124756	0.159842	0.152356	0.159073	0.151493	False	2.30991	0.0173422	2.36194	2.25789	2.3309	2.27314	False
6	[30000:34999]	6	30000	34999			reference	20481.3	271.992	21297.3	19665.3	20978.6	19816.9	False	0.15381	0.00124756	0.157553	0.150068	0.159073	0.151493	False	2.3132	0.0173422	2.36522	2.26117	2.3309	2.27314	False
7	[35000:39999]	7	35000	39999			reference	20657.2	271.992	21473.2	19841.2	20978.6	19816.9	False	0.153576	0.00124756	0.157319	0.149833	0.159073	0.151493	False	2.3062	0.0173422	2.35822	2.25417	2.3309	2.27314	False
8	[40000:44999]	8	40000	44999			reference	20243.4	271.992	21059.4	19427.4	20978.6	19816.9	False	0.156162	0.00124756	0.159904	0.152419	0.159073	0.151493	False	2.30548	0.0173422	2.35751	2.25346	2.3309	2.27314	False
9	[45000:49999]	9	45000	49999			reference	20188.5	271.992	21004.5	19372.5	20978.6	19816.9	False	0.153955	0.00124756	0.157697	0.150212	0.159073	0.151493	False	2.29335	0.0173422	2.34537	2.24132	2.3309	2.27314	False
10	[0:4999]	0	0	4999			analysis	20614.9	271.992	21430.9	19798.9	20978.6	19816.9	False	0.152418	0.00124756	0.156161	0.148675	0.159073	0.151493	False	2.33962	0.0173422	2.39165	2.2876	2.3309	2.27314	True
11	[5000:9999]	1	5000	9999			analysis	20589.5	271.992	21405.5	19773.6	20978.6	19816.9	False	0.155663	0.00124756	0.159405	0.15192	0.159073	0.151493	False	2.30781	0.0173422	2.35984	2.25579	2.3309	2.27314	False
12	[10000:14999]	2	10000	14999			analysis	20463.2	271.992	21279.2	19647.2	20978.6	19816.9	False	0.154717	0.00124756	0.158459	0.150974	0.159073	0.151493	False	2.30841	0.0173422	2.36044	2.25639	2.3309	2.27314	False
13	[15000:19999]	3	15000	19999			analysis	20667	271.992	21483	19851	20978.6	19816.9	False	0.15608	0.00124756	0.159823	0.152337	0.159073	0.151493	False	2.31285	0.0173422	2.36488	2.26083	2.3309	2.27314	False
14	[20000:24999]	4	20000	24999			analysis	19758.5	271.992	20574.5	18942.5	20978.6	19816.9	True	0.153575	0.00124756	0.157318	0.149832	0.159073	0.151493	False	2.31019	0.0173422	2.36222	2.25817	2.3309	2.27314	False
15	[25000:29999]	5	25000	29999			analysis	21804.3	271.992	22620.3	20988.4	20978.6	19816.9	True	0.155871	0.00124756	0.159613	0.152128	0.159073	0.151493	False	2.31176	0.0173422	2.36379	2.25974	2.3309	2.27314	False
16	[30000:34999]	6	30000	34999			analysis	22160.7	271.992	22976.6	21344.7	20978.6	19816.9	True	0.155253	0.00124756	0.158995	0.15151	0.159073	0.151493	False	2.31126	0.0173422	2.36329	2.25923	2.3309	2.27314	False
17	[35000:39999]	7	35000	39999			analysis	21644.4	271.992	22460.4	20828.4	20978.6	19816.9	True	0.155762	0.00124756	0.159505	0.152019	0.159073	0.151493	False	2.31125	0.0173422	2.36328	2.25923	2.3309	2.27314	False
18	[40000:44999]	8	40000	44999			analysis	22013.2	271.992	22829.1	21197.2	20978.6	19816.9	True	0.156886	0.00124756	0.160629	0.153143	0.159073	0.151493	False	2.31088	0.0173422	2.36291	2.25885	2.3309	2.27314	False
19	[45000:49999]	9	45000	49999			analysis	22013.7	271.992	22829.7	21197.7	20978.6	19816.9	True	0.155866	0.00124756	0.159609	0.152123	0.159073	0.151493	False	2.30833	0.0173422	2.36035	2.2563	2.3309	2.27314	False

More information on accessing the information contained in the Result can be found on the Working with results page.

The next step is visualizing the results, which is done using the plot() method. It is recommended to filter results for each column and plot separately.

>>> for column_name in results.column_names:
...     results.filter(column_names=column_name).plot().show()

Insights

We see that only the car_value column exhibits a permanent change in standard deviation values. However both car_value and driver_tenure appear to have one off events where they have some slightly abnormal values.

What Next

We can also inspect the dataset for other Summary Statistics such as Average. We can also look for any Data Drift present in the dataset using Detecting Data Drift functionality of NannyML.