PyCaret 3.3.2

Release: PyCaret 3.3.2

• Fixes installation and AUC metric (as well as other probability based metric) issues.

What's Changed

• Update memory.py - removing joblib._format_load_msg() usage by @hakan-77 in #3980
• Fix pycaret[full] cannot be installed by @jmakov in #3982
• Fix AUC metric by @celestinoxp in #3935

New Contributors

• @hakan-77 made their first contribution in #3980
• @jmakov made their first contribution in #3982

Full Changelog: 3.3.1...3.3.2

PyCaret 3.3.1

Release: PyCaret 3.3.1

• Maintenance update to fix dependency and compatibility issues.

What's Changed

• Add files via upload by @1jiangxd in #3858
• Edit portuguese translation in tutorials/translations/portuguese by @Laura-Sangalli in #3903
• Upgrade old evidently api to new api by @celestinoxp in #3923
• Custom path for logging (logs.log) by @rhsouza in #3902
• Bump pyyaml from 5.3.1 to 5.4 by @dependabot in #3911
• fix ransac loss by @celestinoxp in #3920
• Fix AUC metrics by @celestinoxp in #3926
• Fix dependency: Joblib 1.4.0 breaks PyCaret, so downgrade to <1.4 by @amotl in #3967
• Fix sklearnex, shap and dask CI by @Yard1 in #3968

New Contributors

• @1jiangxd made their first contribution in #3858
• @Laura-Sangalli made their first contribution in #3903
• @rhsouza made their first contribution in #3902
• @dependabot made their first contribution in #3911

Full Changelog: 3.3.0...3.3.1

PyCaret 3.3.0

Release: PyCaret 3.3.0

• Support for latest version of scikit-learn (1.4) and pandas (2.0) (special thanks to @celestinoxp)
• Python 3.8 support has been dropped!

Changes

• Remove badge of python 3.7 by @celestinoxp in #3831
• Support Pandas 2.0 by @celestinoxp in #3830
• fix moving categorical with 2 unique values to ordinal by @tvdboom in #3848
• bump "triad" version from 0.9.1 to 0.9.3 (requirements-optional.txt) by @celestinoxp in #3828
• Fix create_docker() by @rafinhaLQ in #3838
• fixed issue where the "return_train_score" parameter is not passed by @Nelsaur in #3840
• (#3841) raise supported MLflow version by @pascal456 in #3842
• (#3870) align default value for restack in stack_models by @pascal456 in #3871
• Update copyright headers by @celestinoxp in #3892
• Support Scikit-learn 1.4 by @celestinoxp in #3857

New Contributors

• @rafinhaLQ made their first contribution in #3838
• @Nelsaur made their first contribution in #3840

Full Changelog: 3.2.0...3.3.0

PyCaret 3.2.0

Release: PyCaret 3.2.0 (BUG FIXES)

• Python 3.11 support has been added!

Changes

• During the process of setting up the installation environment, upgrad… by @turkalpmd in #3658
• Fix MLflowClient import by @celestinoxp in #3745
• [docs] update LightGBM GPU installation docs by @jameslamb in #3734
• fix index handling with iterative imputer by @tvdboom in #3762
• Update README.md by @moezali1 in #3753
• fix error raise in compare_models by @tvdboom in #3773
• Pin Werkzeug>=2.2,<3.0 to fix Flask by @Yard1 in #3800
• switched to non-positional explicit args and kwargs to catch them by @ngupta23 in #3801
• Doc(Update version): No longer support python 3.7 by @wjrforcyber in #3771
• Fix target transformation metrics by @Yard1 in #3799
• Dependencies: Permit NumPy up to 1.26. CI: Run Python 3.11 on test_linux workflow. by @amotl in #3813
• [BUG]: default value for optimize in automl() in regression modules has to be changed by @pascal456 in #3791
• (#3796) fix index reference in API creation by @pascal456 in #3817
• More Python 3.11 support by @amotl in #3814
• Optimize CI by @amotl in #3815
• Update xgboost v2 and restrict Python versions by @celestinoxp in #3749
• add python 3.11 support by @tvdboom in #3756
• Update init.py by @celestinoxp in #3819

New Contributors

• @turkalpmd made their first contribution in #3658
• @jameslamb made their first contribution in #3734
• @wjrforcyber made their first contribution in #3771
• @amotl made their first contribution in #3813
• @pascal456 made their first contribution in #3791

Full Changelog: 3.1.0...3.2.0

PyCaret 3.1.0

Release: PyCaret 3.1.0 (BUG FIXES, API CHANGES)

• Python 3.7 support has been dropped and PyCaret is no longer tested with Python 3.7 or guaranteed to work with 3.7.
• deep_check and eda functions have been removed. Calling those functions will raise exceptions.

Changes

• Fix groups and pipelines that alter number of rows by @tvdboom in #3629
• Fixed dead links in Tutorial docs by @mbellitti in #3646
• Fix for sktime changes in recent versions by @ngupta23 in #3652
• [Bug Fix] Fixed issues caused by old deprecated pandas-profiling library to new updated ydata-profiling by @beingmechon in #3649
• removed support for python 3.7 by @ngupta23 in #3677
• [BUG]: Enable random_state for class imbalance by @arghhjayy in #3676
• Update README.md by @moezali1 in #3682
• Resolve double-printed plots during ts experiment by @chrimaho in #3670
• fix groups bug, cuml version bug and missing installed libraries bug by @tvdboom in #3710
• limiting statsforecast due to deprecation of parallel argument by @ngupta23 in #3731
• Validate data_split_shuffle for timeseries CV by @Yard1 in #3711
• Fix classification custom metric label encoding by @Yard1 in #3715
• Fix iterative imputer with many categories by @Yard1 in #3713
• Strip pipeline in tune_model, fix memory pickling by @Yard1 in #3714
• Remove deepchecks and eda by @Yard1 in #3712

New Contributors

• @mbellitti made their first contribution in #3646
• @beingmechon made their first contribution in #3649
• @arghhjayy made their first contribution in #3676

Full Changelog: 3.0.4...3.1.0

PyCaret 3.0.4

Release: PyCaret 3.0.4 (MAINTENANCE UPDATE)

Changes

• Pin sklearn<1.3.0, bump version to 3.0.4, cleanup by @Yard1 in #3628

Full Changelog: 3.0.3...3.0.4

PyCaret 3.0.3

Release: PyCaret 3.0.3 (BUG FIXES)

Changes

• fix feature selection by @tvdboom in #3562
• Remove deprecated matplotlib parameter by @Yard1 in #3574
• DagsHubLogger: support programmatic logging / Bug Fix: incorrect url splitter, mlflow nested run by @timho102003 in #3577
• Update README.md by @moezali1 in #3584
• Fix missing USI argument in CometLogger by @Yard1 in #3587
• Pin catboost<1.2 on Mac by @Yard1 in #3586
• Fixes TS Plotting Issue by @ngupta23 in #3599
• prep for pipeline change needed when we upgrade sktime to 0.19.1 by @ngupta23 in #3598
• Version related fixes by @ngupta23 in #3596
• Remove use_train_data from plot_model by @Yard1 in #3595
• Fix bad label dtype in clf predict_model by @Yard1 in #3619
• Fix column reorder causing duplicate columns by @Yard1 in #3620
• Force np.float64 for CBLOF by @Yard1 in #3621

New Contributors

• @timho102003 made their first contribution in #3577

Full Changelog: 3.0.2...3.0.3

PyCaret 3.0.2

Release: PyCaret 3.0.2 (BUG FIXES)

Changes

• Fix MLflow nesting and model artifacts by @Yard1 in #3559
• Unpin holoviews by @Yard1 in #3560

Full Changelog: 3.0.1...3.0.2

PyCaret 3.0.1

Release: PyCaret 3.0.1 (BUG FIXES)

Changes

• Added and organized badges by @ngupta23 in #3412
• Pin numpy<1.24 by @Yard1 in #3414
• Cleanup cuML ImportError message by @beckernick in #3423
• add duplicate indices check by @tvdboom in #3430
• added index cleaning utility for time series by @ngupta23 in #3448
• fix of CatboostRegresser on GPU with distribution. by @kyosuke0924 in #3458
• Comet Compare Logging Bug Fix by @sherpan in #3462
• Restrict evidently version by @mike0sv in #3465
• fix np array check when get_feature_names_out is unavailable by @tvdboom in #3470
• Update functional.py by @Denny-Hwang in #3473
• pinned sktime until issue is fixed by @ngupta23 in #3474
• disabling failing tests for now. by @ngupta23 in #3494
• added STLForecaster by @ngupta23 in #3476
• Added dynamic seasonal type detection and removed bad models by @ngupta23 in #3477
• update to new schemdraw by @tvdboom in #3500
• Fix for TBATS issues on M3 dataset by @ngupta23 in #3478
• Check duplicate cols by @tvdboom in #3485
• Enhanced seasonality type detection by @ngupta23 in #3518
• Pin holoviews<=1.14.9 in requirements-optional by @Yard1 in #3521
• Fix fugue tests by @goodwanghan in #3504
• Fix predict_model with transform_target by @Yard1 in #3519
• Ensure finalize_model logs correct pipeline by @Yard1 in #3520
• Fix tree plot with no label encoder by @Yard1 in #3480
• provided option for users to provide seasonality type by @ngupta23 in #3533
• Time Series blend_models enhancements and bug fixes by @ngupta23 in #3534
• Fix metrics being zero in predict_model by @Yard1 in #3537
• Fix save_model with catboost by @Yard1 in #3481
• Remove leaveoneout by @tvdboom in #3543

New Contributors

• @kyosuke0924 made their first contribution in #3458
• @mike0sv made their first contribution in #3465
• @Denny-Hwang made their first contribution in #3473

Full Changelog: 3.0.0...3.0.1

PyCaret 3.0

Release: PyCaret 3.0

This is a major update introducing multiple new features and bug fixes as well as breaking API changes. Please refer to the documentation and tutorials for more information.

PyCaret 2.0 code will most likely not run without changes!

Main changes

• New Time Series Forecasting module
• Far-reaching internal refactor for better performance, consistency and developer productivity
• New object-oriented API that can be used alongside the functional API
• Completely refactored preprocessing
• Prevented target leakage by using pipelines in cross-validation
• Compatibility with latest scikit-learn
• Support for scikit-learn-intelex for better performance on CPU
• New experiment logging API with MLFLow, Weights & Biases, DagsHub and CometML support
• Distributed parallel computation with Fugue
• Reworked tutorials and documentation
• Removal of outdated NLP and ARules modules
• Multiple other fixes and improvements

Full Changelog

• Black by @Yard1 in #836
• Black by @Yard1 in #838
• Added unit tests for simple imputer and fixed minor issues by @wkuopt in #974
• Added imputation strategy for time columns for simple imputer by @wkuopt in #978
• Fixed the cutoff of long text in png graphs by @wkuopt in #992
• Fixed Dummify for satisfying sklearn pipeline structure by @wkuopt in #1004
• added initial dataset per #1038 by @ngupta23 in #1044
• Time series metrics by @TremaMiguel in #1067
• Fixed Reduce_Dimensions_For_Supervised_Path for satisfying sklearn pipeline by @wkuopt in #1006
• Time series splitter by @TremaMiguel in #1082
• Time Series (create_model) - Initial PR by @ngupta23 in #1068
• Added more model types by @ngupta23 in #1089
• Time series ensemble models by @TremaMiguel in #1105
• tune_model + create_model updates by @ngupta23 in #1114
• Time Series metrics cleanup by @ngupta23 in #1119
• Time series seasonal parameter and fh splitting data. by @TremaMiguel in #1129
• Initial additions for Random Grid Search for Time Series by @ngupta23 in #1164
• Better Grid Search by @ngupta23 in #1173
• feat: compare time series models by @TremaMiguel in #1174
• Break up PycaretExperiment by @Yard1 in #1175
• Fixes circular imports by @ngupta23 in #1177
• Alternative approach to TS scikit models by @Yard1 in #1196
• added more scikit models along with proper grid search (previously placeholder only) by @ngupta23 in #1195
• Seasonality check Flag by @TremaMiguel in #1199
• Ts scikit models by @ngupta23 in #1198
• Improve create model + change seasonal_parameter to seasonal_period by @ngupta23 in #1210
• feat: Tbats and bats models by @TremaMiguel in #1243
• improved defaults and improved flow demonstration by @ngupta23 in #1242
• Ts datasets by @ngupta23 in #1247
• Ts moez feedback by @ngupta23 in #1256
• Cleaning up models by @ngupta23 in #1265
• Fold Strategy Updates by @ngupta23 in #1282
• Added catboost model by @ngupta23 in #1286
• [FEAT] Auto ARIMA by @ngupta23 in #1294
• tune_model fixes and cleanup by @ngupta23 in #1306
• Time Series predict_model by @ngupta23 in #1332
• finalize_model by @ngupta23 in #1340
• added dummylogger by @tvdboom in #1341
• Support for negative values in time series by @ngupta23 in #1347
• Support for custom folds by @ngupta23 in #1353
• fix for non default optimize metric by @ngupta23 in #1361
• added display for test metrics in predict model by @ngupta23 in #1363
• Added cutoffs to score dataframe by @ngupta23 in #1370
• fix: time series model types available by @TremaMiguel in #1354
• fix: Blend model voting with weight Uniforms by @TremaMiguel in #1375
• fix for compare_model folds by @ngupta23 in #1378
• Added R2 metric container by @ngupta23 in #1380
• feat: Prophet model by @TremaMiguel in #1279
• Add polynomial_features to preprocessing by @IncubatorShokuhou in #1434
• initial changes until splitters by @ngupta23 in #1462
• more updates to make sktime 0.7.0 compatible. Disabled ensemble/blend for now by @ngupta23 in #1466
• fix: blending compatible with sktime 0.7.0 by @TremaMiguel in #1469
• Add transformation by @IncubatorShokuhou in #1451
• Ts sktime 0p7 by @ngupta23 in #1493
• Added test_model functionality by @ngupta23 in #1495
• [timeseries] Fix_fit_and_score metrics not getting their kwargs by @Yard1 in #1512
• Use __getattr__ in pipeline for fit vars by @Yard1 in #1511
• Added Time Series Plots by @satya-pattnaik in #1544
• plot_model functionality by @ngupta23 in #1513
• Multiple Time Series Fixes and Enhancements by @ngupta23 in #1585
• Update Plots by @ngupta23 in #1587
• Time series fix #1558, #1556, bad assert in compare_models, etc. by @Yard1 in #1586
• Ts plot modifications by @satya-pattnaik in #1614
• Further fixes to time series module by @ngupta23 in #1615
• Fix load_model() not working without running setup() first by @Yard1 in #1617
• Fix Stats Index by @ngupta23 in #1619
• Multiple Fixes and Residual Analysis Added by @ngupta23 in #1620
• Pycaret time series - Added Decomp Plots(classical/STL) by @satya-pattnaik in #1657
• Added decomposition plot by @satya-pattnaik in #1647
• added decomp wrapper per #1580 and cleanup by @ngupta23 in #1661
• Option to enforce Predicton Interval by @ngupta23 in #1662
• chanegs per #1663 by @ngupta23 in #1664
• Time Series Bug Fixes and Error Handling by @ngupta23 in #1668
• Fix MLFlow logging for Time Series by @Yard1 in #1673
• Refactor unit tests by @ngupta23 in #1687
• Time Series fixes before Beta Release by @ngupta23 in #1704
• Datasets + Plot Model fixes by @ngupta23 in #1712
• master to time_series by @Yard1 in #1706
• Fix mlflow params logging by @Yard1 in #1718
• fix added for #1713 which broke because of prev fix for #1702 by @ngupta23 in #1719
• update docstrings for pycaret.time_series by @moezali1 in #1717
• more updates to time_series docstring by @moezali1 in #1729
• Update requirements.txt by @Yard1 in #1733
• Update requirements.txt by @Yard1 in #1734
• Fix log_plots functionality by @Yard1 in #1737
• Pin statsmodels, plotly by @Yard1 in #1738
• TS beta fixes by @ngupta23 in #1742
• Time series Beta Release updates by @ngupta23 in #1743
• Fix custom grids for time series by @Yard1 in #1776
• Ts fixes by @ngupta23 in #1774
• Time series fixes to push to beta rel...