Deepchecks - Test Suites for ML Models and Data
Deepchecks is a Python package for comprehensively validating your machine learning models and data with minimal effort. This includes checks related to various types of issues, such as model performance, data integrity, distribution mismatches, and more.
Key Concepts
Check
Each check enables you to inspect a specific aspect of your data and models. They are the basic building block of the deepchecks package, covering all kinds of common issues, such as: PerformanceOverfit, DataSampleLeakage, SingleFeatureContribution, DataDuplicates, and many more checks. Each check can have two types of results:
- A visual result meant for display (e.g. a figure or a table).
- A return value that can be used for validating the expected check results (validations are typically done by adding a "condition" to the check, as explained below).
Condition
A condition is a function that can be added to a Check, which returns a pass ✓, fail
from deepchecks.checks import BoostingOverfit
BoostingOverfit().add_condition_test_score_percent_decline_not_greater_than(threshold=0.05)
which will fail if there is a difference of more than 5% between the best score achieved on the test set during the boosting iterations and the score achieved in the last iteration (the model's "original" score on the test set).
Suite
An ordered collection of checks, that can have conditions added to them. The Suite enables displaying a concluding report for all of the Checks that ran. Here you can find the predefined existing suites and a code example demonstrating how to build your own custom suite. The existing suites include default conditions added for most of the checks. You can edit the preconfigured suites or build a suite of your own with a collection of checks and optional conditions.
Installation
Using pip
pip install deepchecks #--user
From source
First clone the repository and then install the package from inside the repository's directory:
git clone https://github.com/deepchecks/deepchecks.git
cd deepchecks
# for installing stable tag version and not the latest commit to main
git checkout tags/<version>
and then either:
pip install .
or
python setup.py install
Are You Ready to Start Checking?
For the full value from Deepchecks' checking suites, we recommend working with:
-
A model compatible with scikit-learn API that you wish to validate (e.g. RandomForest, XGBoost)
-
The model's training data with labels
-
Test data (on which the model wasn’t trained) with labels
However, many of the checks and some of the suites need only a subset of the above to run.
Usage Examples
Running a Check
For running a specific check on your pandas DataFrame, all you need to do is:
from deepchecks.checks import RareFormatDetection
import pandas as pd
df_to_check = pd.read_csv('data_to_validate.csv')
# Initialize and run desired check
RareFormatDetection().run(df_to_check)
Which might product output of the type:
Rare Format Detection
Check whether columns have common formats (e.g. 'XX-XX-XXXX' for dates) and detects values that don't match.
✓ Nothing found
If all was fine, or alternatively something like:
Rare Format Detection
Check whether columns have common formats (e.g. 'XX-XX-XXXX' for dates) and detects values that don't match.
Column date:
digits and letters format (case sensitive) ratio of rare samples 1.50% (3) common formats ['2020-00-00'] examples for values in common formats ['2021-11-07'] values in rare formats ['2021-Nov-04', '2021-Nov-05', '2021-Nov-06']
If mismatches were detected.
Running a Suite
Let's take the "iris" dataset as an example:
import pandas as pd
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
iris_df = load_iris(return_X_y=False, as_frame=True)['frame']
label_col = 'target'
df_train, df_test = train_test_split(iris_df, stratify=iris_df[label_col], random_state=0)
To run an existing suite all you need to do is import the suite and run it -
from deepchecks.suites import integrity_check_suite
integrity_suite = integrity_check_suite()
integrity_suite.run(train_dataset=df_train, test_dataset=df_test, check_datasets_policy='both')
Which will result in printing the summary of the check conditions and then the visual outputs of all of the checks that are in that suite.
Example Notebooks
For usage examples, check out:
- deepchecks Quick Start Notebook - for a simple example notebook for working with checks and suites.
- Example Checks Output Notebooks - to see all of the existing checks and their usage examples.
Communication
- Join our Slack Community to connect with the maintainers and follow users and interesting discussions
- Post a Github Issue to suggest improvements, open an issue, or share feedback.