Skip to content
forked from NannyML/nannyml

Detecting silent model failure. NannyML estimates performance with an algorithm called Confidence-based Performance estimation (CBPE), developed by core contributors. It is the only open-source algorithm capable of fully capturing the impact of data drift on performance.

License

Notifications You must be signed in to change notification settings

NeoKish/nannyml

ย 
ย 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 

Repository files navigation

Documentation Status PyPI - License

NannyML - OSS Python library for detecting silent ML model failure | Product Hunt

Website โ€ข Docs โ€ข Community Slack

animated

๐Ÿ’ก What is NannyML?

NannyML is an open-source python library that allows you to estimate post-deployment model performance (without access to targets), detect data drift, and intelligently link data drift alerts back to changes in model performance. Built for data scientists, NannyML has an easy-to-use interface, interactive visualizations, is completely model-agnostic and currently supports all tabular use cases, classification and regression.

The core contributors of NannyML have researched and developed multiple novel algorithms for estimating model performance: confidence-based performance estimation (CBPE) and direct loss estimation (DLE). The nansters also invented a new approach to detect multivariate data drift using PCA-based data reconstruction.

If you like what we are working on, be sure to become a Nanster yourself, join our community slack and support us with a GitHub star โญ.

โ˜” Why use NannyML?

NannyML closes the loop with performance monitoring and post deployment data science, empowering data scientist to quickly understand and automatically detect silent model failure. By using NannyML, data scientists can finally maintain complete visibility and trust in their deployed machine learning models. Allowing you to have the following benefits:

  • End sleepless nights caused by not knowing your model performance ๐Ÿ˜ด
  • Analyse data drift and model performance over time
  • Discover the root cause to why your models are not performing as expected
  • No alert fatigue! React only when necessary if model performance is impacted
  • Painless setup in any environment

๐Ÿง  GO DEEP

NannyML Resources Description
โ˜Ž๏ธ NannyML 101 New to NannyML? Start here!
๐Ÿ”ฎ Performance estimation How the magic works.
๐ŸŒ Real world example Take a look at a real-world example of NannyML.
๐Ÿ”‘ Key concepts Glossary of key concepts we use.
๐Ÿ”ฌ Technical reference Monitor the performance of your ML models.
๐Ÿ”Ž Blog Thoughts on post-deployment data science from the NannyML team.
๐Ÿ“ฌ Newsletter All things post-deployment data science. Subscribe to see the latest papers and blogs.
๐Ÿ’Ž New in v0.10.6 New features, bug fixes.
๐Ÿง‘โ€๐Ÿ’ป Contribute How to contribute to the NannyML project and codebase.
Join slack Need help with your specific use case? Say hi on slack!

๐Ÿ”ฑ Features

1. Performance estimation and monitoring

When the actual outcome of your deployed prediction models is delayed, or even when post-deployment target labels are completely absent, you can use NannyML's CBPE-algorithm to estimate model performance for classification or NannyML's DLE-algorithm for regression. These algorithms provide you with any estimated metric you would like, i.e. ROC AUC or RSME. Rather than estimating the performance of future model predictions, CBPE and DLE estimate the expected model performance of the predictions made at inference time.

NannyML can also track the realised performance of your machine learning model once targets are available.

2. Data drift detection

To detect multivariate feature drift NannyML uses PCA-based data reconstruction. Changes in the resulting reconstruction error are monitored over time and data drift alerts are logged when the reconstruction error in a certain period exceeds a threshold. This threshold is calculated based on the reconstruction error observed in the reference period.

NannyML utilises statistical tests to detect univariate feature drift. We have just added a bunch of new univariate tests including Jensen-Shannon Distance and L-Infinity Distance, check out the comprehensive list. The results of these tests are tracked over time, properly corrected to counteract multiplicity and overlayed on the temporal feature distributions. (It is also possible to visualise the test-statistics over time, to get a notion of the drift magnitude.)

NannyML uses the same statistical tests to detected model output drift.

Target distribution drift can also be monitored using the same statistical tests. Bear in mind that this operation requires the presence of actuals.

3. Intelligent alerting

Because NannyML can estimate performance, it is possible to weed out data drift alerts that do not impact expected performance, combatting alert fatigue. Besides linking data drift issues to drops in performance it is also possible to prioritise alerts according to other criteria using NannyML's Ranker.

๐Ÿš€ Getting started

Install NannyML

NannyML depends on LightGBM. This might require you to set install additional OS-specific binaries. You can follow the official installation guide.

From PyPI:

pip install nannyml

From Conda:

 conda install -c conda-forge nannyml

Running via Docker:

docker -v /local/config/dir/:/config/ run nannyml/nannyml nml run

Here be dragons! Use the latest development version of NannyML at your own risk:

python -m pip install git+https://github.com/NannyML/nannyml

Quick Start

The following snippet is based on our latest release.

import nannyml as nml
import pandas as pd
from IPython.display import display

# Load real-world data:
reference_df, analysis_df, _ = nml.load_us_census_ma_employment_data()
display(reference_df.head())
display(analysis_df.head())

# Choose a chunker or set a chunk size:
chunk_size = 5000

# initialize, specify required data columns, fit estimator and estimate:
estimator = nml.CBPE(
    problem_type='classification_binary',
    y_pred_proba='predicted_probability',
    y_pred='prediction',
    y_true='employed',
    metrics=['roc_auc'],
    chunk_size=chunk_size,
)
estimator = estimator.fit(reference_df)
estimated_performance = estimator.estimate(analysis_df)

# Show results:
figure = estimated_performance.plot()
figure.show()

# Define feature columns:
features = ['AGEP', 'SCHL', 'MAR', 'RELP', 'DIS', 'ESP', 'CIT', 'MIG', 'MIL', 'ANC',
       'NATIVITY', 'DEAR', 'DEYE', 'DREM', 'SEX', 'RAC1P']

# Initialize the object that will perform the Univariate Drift calculations:
univariate_calculator = nml.UnivariateDriftCalculator(
    column_names=features,
    chunk_size=chunk_size
)

univariate_calculator.fit(reference_df)
univariate_drift = univariate_calculator.calculate(analysis_df)

# Get features that drift the most with count-based ranker:
alert_count_ranker = nml.AlertCountRanker()
alert_count_ranked_features = alert_count_ranker.rank(univariate_drift)
display(alert_count_ranked_features.head())

# Plot drift results for top 3 features:
figure = univariate_drift.filter(column_names=['RELP','AGEP', 'SCHL']).plot()
figure.show()

# Compare drift of a selected feature with estimated performance
uni_drift_AGEP_analysis = univariate_drift.filter(column_names=['AGEP'], period='analysis')
figure = estimated_performance.compare(uni_drift_AGEP_analysis).plot()
figure.show()

# Plot distribution changes of the selected features:
figure = univariate_drift.filter(period='analysis', column_names=['RELP','AGEP', 'SCHL']).plot(kind='distribution')
figure.show()

# Get target data, calculate, plot and compare realized performance with estimated performance:
_, _, analysis_targets_df = nml.load_us_census_ma_employment_data()

analysis_with_targets_df = pd.concat([analysis_df, analysis_targets_df], axis=1)
display(analysis_with_targets_df.head())

performance_calculator = nml.PerformanceCalculator(
    problem_type='classification_binary',
    y_pred_proba='predicted_probability',
    y_pred='prediction',
    y_true='employed',
    metrics=['roc_auc'],
    chunk_size=chunk_size)

performance_calculator.fit(reference_df)
calculated_performance = performance_calculator.calculate(analysis_with_targets_df)

figure = estimated_performance.filter(period='analysis').compare(calculated_performance).plot()
figure.show()

๐Ÿ“– Documentation

๐Ÿฆธ Contributing and Community

We want to build NannyML together with the community! The easiest to contribute at the moment is to propose new features or log bugs under issues. For more information, have a look at how to contribute.

๐Ÿ™‹ Get help

The best place to ask for help is in the community slack. Feel free to join and ask questions or raise issues. Someone will definitely respond to you.

๐Ÿฅท Stay updated

If you want to stay up to date with recent changes to the NannyML library, you can subscribe to our release notes. For thoughts on post-deployment data science from the NannyML team, feel free to visit our blog. You can also sing up for our newsletter, which brings together the best papers, articles, news, and open-source libraries highlighting the ML challenges after deployment.

๐Ÿ“ Roadmap

Curious what we are working on next? Have a look at our roadmap. If you have any questions or if you would like to see things prioritised in a different way, let us know!

๐Ÿ“ Citing NannyML

To cite NannyML in academic papers, please use the following BibTeX entry.

Version 0.10.6

    @misc{nannyml,
        title = {{N}anny{ML} (release 0.10.6)},
        howpublished = {\url{https://github.com/NannyML/nannyml}},
        month = mar,
        year = 2023,
        note = {NannyML, Belgium, OHL.},
        key = {NannyML}
    }

๐Ÿ“„ License

NannyML is distributed under an Apache License Version 2.0. A complete version can be found here. All contributions will be distributed under this license.

About

Detecting silent model failure. NannyML estimates performance with an algorithm called Confidence-based Performance estimation (CBPE), developed by core contributors. It is the only open-source algorithm capable of fully capturing the impact of data drift on performance.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 99.7%
  • Other 0.3%