Skip to content

Latest commit

 

History

History
700 lines (519 loc) · 53.3 KB

CHANGELOG.md

File metadata and controls

700 lines (519 loc) · 53.3 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

Release: PyCaret 2.3.3 | Release Date: July 24th, 2021 (NEW FEATURES, BUG FIXES)

  • Fixed issues with [full] install by pinning interpret<=0.2.4
  • Added support for S3 folder path in deploy_model() with AWS
  • Enabled experimental Optuna TPESampler options to improve convergence (in tune_model())


Release: PyCaret 2.3.2 | Release Date: July 7th, 2021 (NEW FEATURES, BUG FIXES)

  • Implemented PDP, MSA and PFI plots in interpret_model - thanks to @IncubatorShokuhou (pycaret#1415)
  • Implemented Kolmogorov-Smirnov (KS) plot in plot_model under pycaret.classification module
  • Fixed a typo "RVF" to "RBF" - thanks to @baturayo (pycaret#1220)
  • Readme & license updates and improvements
  • Fixed remove_multicollinearity considering categorical features
  • Fixed keyword issues with PyCaret's cuML wrappers
  • Improved performance of iterative imputation
  • Fixed gain and lift plots taking wrong arguments, creating misleading plots
  • interpret_model on LightGBM will now show a beeswarm plot
  • Multiple improvements to exception handling and documentation in pycaret.persistence (pycaret#1324)
  • remove_perfect_collinearity option will now be show in the setup() summary - thanks to @mjkanji (pycaret#1342)
  • Fixed IterativeImputer setting wrong float precision
  • Fixed custom grids in tune_model raising an exception when composed of lists
  • Improved documentation in pycaret.clustering - thanks to @susmitpy (pycaret#1372)
  • Added support for LightGBM CUDA version - thanks to @IncubatorShokuhou (pycaret#1396)
  • Exposed address in get_data for alternative data sources - thanks to @IncubatorShokuhou (pycaret#1416)


Release: PyCaret 2.3.1 | Release Date: April 28, 2021 (SEVERAL BUGS FIXED)

  • Fixed an exception with missing variables (display_container etc.) during load_config()
  • Fixed exceptions when using Ridge and RF estimators with cuML (GPU mode)
  • Fixed PyCaret's cuML wrappers not being pickleable
  • Added an extra check to get_all_object_vars_and_properties internal method, fixing exceptions with certain estimators
  • save_model() now supports kwargs, which will be passed to joblib.dump()
  • Fixed an issue with load_model() from AWS (duplicate .pkl extension) - thanks to markgrujic (pycaret#1128)
  • Fixed a typo in documentation - thanks to koorukuroo (pycaret#1149)
  • Optimized Fix_multicollinearity transformer, drastically reducing the size of saved pipeline
  • interpret_model() now supports data passed as an argument - thanks to jbechtel (pycaret#1184)
  • Removed infer_signature from MLflow logging when log_experiment=True.
  • Fixed a rare issue where binary_multiclass_score_func was not pickleable
  • Fixed edge case exceptions in feature selection
  • Fixed an exception with finalize_model when using GroupKFold CV
  • Pinned mlxtend>=0.17.0, imbalanced-learn==0.7.0, and gensim<4.0.0


Release: PyCaret 2.3.0 | Release Date: February 21, 2021

  • Modules Impacted: pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.arules

Summary of Changes

  • Added new interactive residual plots in pycaret.regression module. You can now generate interactive residual plots by using residuals_interactive in the plot_model function.
  • Added plot rendering support for streamlit applications. A new parameter display_format is added in the plot_model function. To render plot in streamlit app, set this to streamlit.
  • Revamped Boruta feature selection algorithm. (give it a try!).
  • tune_model in pycaret.classification and pycaret.regression is now compatible with custom models.
  • Added low_memory and max_len support to association rules module (pycaret#1008).
  • Increased robustness of DataFrame checks (pycaret#1005).
  • Improved loading of models from AWS (pycaret#1005).
  • Catboost and XGBoost are now optional dependencies. They are not automatically installed with default slim installation. To install optional dependencies use pip install pycaret[full].
  • Added raw_score argument in the predict_model function for pycaret.classification module. When set to True, scores for each class will be returned separately.
  • PyCaret now returns base scikit-learn objects, whenever possible.
  • When handle_unknown_categorical is set to False in the setup function, an exception will be raised during prediction if the data contains unknown levels in categorical features.
  • predict_model for multiclass classification now returns labels as an integer.
  • Fixed an edge case where an IndexError would be raised in pycaret.clustering and pycaret.anomaly.
  • Fixed text formatting for certain plots in pycaret.classification and pycaret.regression.
  • If a logs.log file cannot be created when setup is initialized, no exception will be raised now (support for more configurable logging to come in future).
  • User added metrics will not raise exceptions now and instead return 0.0.
  • Compatibility with tune-sklearn>=0.2.0.
  • Fixed an edge case for dropping NaNs in target column.
  • Fixed stacked models not being tuned correctly.
  • Fixed an exception with KFold when fold_shuffle=False.


Release: PyCaret 2.2.3 | Release Date: December 22, 2020 (SEVERAL BUGS FIX | CRITICAL COMPATIBILITY FIX)

  • Fixed exceptions with the predict_model function when data columns had non-string characters.
  • Fixed a rare exception with the remove_multicollinearity parameter in the setup function`.
  • Improved performance and robustness of conversion of date features to categoricals.
  • Fixed an exception with the models function when the type parameter was passed.
  • The data frame displayed after setup can now be accessed with the pull function.
  • Fixed an exception with save_config
  • Fixed a rare case where the target column would be treated as an ID column and thus dropped.
  • SHAP plots can now be saved (pass save parameter as True)
  • | CRITICAL | Compatibility broke for catboost, pyod (other impacts unknown as of now) with sklearn=0.24 (released on Dec 22, 2020). A temporary fix is requiring 0.23.2 specifically in the requirements.txt.


Release: PyCaret 2.2.2 | Release Date: November 25, 2020 (SEVERAL BUGS FIX)

  • Fixed an issue with the optimize_threshold function the pycaret.classification module. It now returns a float instead of an array.
  • Fixed issue with the predict_model function. It now uses original data frame to append the predictions. As such any extra columns given at the time of inference are not removed when returning the predictions. Instead they are internally ignored at the time of predictions.
  • Fixed edge case exceptions for the create_model function in pycaret.clustering.
  • Fixed exceptions when column names are not string.
  • Fixed exceptions in pycaret.regression when transform_target is True in the setup function.
  • Fixed an exception in the models function if the type parameter is specified.


Release: PyCaret 2.2.1 | Release Date: November 09, 2020 (SEVERAL BUGS FIX)

Post-release 2.2, the following issues have been fixed:

  • Fixed plot_model = 'tree' exceptions.
  • Fixed issue with predict_model causing errors with non-contiguous indices.
  • Fixed issue with remove_outliers parameter in the setup function. It was introducing extra columns in training data. The issue has been fixed now.
  • Fixed issue with plot_model in pycaret.clustering causing errors with non-contiguous indices.
  • Fixed an exception when the model was saved or logged when imputation_type is set to 'iterative' in the setup function.
  • compare_models now prints intermediate output when html=False.
  • Metrics in pycaret.classification for binary classification are now calculated with average='binary'. Before they were a weighted average of positive and negative class, now they are just calculated for positive class. For multiclass classification average='weighted'.
  • optimize_threshold now returns optimized probability threshold value as numpy object.
  • Fixed issue with certain exceptions in compare_models.
  • Added profile_kwargs argument in the setup function to pass keyword arguments to Pandas Profiler.
  • plot_model, interpret_model, and evaluate_model now accepts a new parameter use_train_data which when set to True, generates plot on train data instead of test data.


Release: PyCaret 2.2 | Release Date: October 28, 2020

Summary of Changes

  • Modules Impacted: pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly

  • Separate Train and Test Set: New parameter test_data has been added in the setup function of pycaret.classification and pycaret.regression. When a DataFrame is passed into the test_data, it is used as a holdout set and the train_size parameter is ignored. test_data must be labeled and the shape of test_data must match with the shape of data.

  • Disable Default Preprocessing: A new parameter preprocess has been added into the setup function. When preprocess is set to False, no transformations are applied except for train_test_split and custom transformations passed in the custom_pipeline param. Data must be ready for modeling (no missing values, no dates, categorical data encoding) when preprocess is set to False.

  • Custom Metrics: New functions get_metric, add_metric and remove_metric is now added in pycaret.classification, pycaret.regression, and pycaret.clustering, that can be used to add / remove metrics used in model evaluation.

  • Custom Transformations: A new parameter custom_pipeline has been added into the setup function. It takes a tuple of (str, transformer) or a list of tuples. When passed, it will append the custom transformers in the preprocessing pipeline and are applied on each CV fold separately and on the final fit. All the custom transformations are applied after train_test_split and before pycaret's internal transformations.

  • GPU enabled Training: To use GPU for training use_gpu parameter in the setup function can be set to True or force. When set to True, it will use GPU with algorithms that support it and fall back on CPU for remaining. When set to force it will only use GPU-enabled algorithms and raise exceptions if they are unavailable for use. The following algorithms are supported on GPU:

    • Extreme Gradient Boosting pycaret.classification pycaret.regression
    • LightGBM pycaret.classification pycaret.regression
    • CatBoost pycaret.classification pycaret.regression
    • Random Forest pycaret.classification pycaret.regression
    • K-Nearest Neighbors pycaret.classification pycaret.regression
    • Support Vector Machine pycaret.classification pycaret.regression
    • Logistic Regression pycaret.classification
    • Ridge Classifier pycaret.classification
    • Linear Regression pycaret.regression
    • Lasso Regression pycaret.regression
    • Ridge Regression pycaret.regression
    • Elastic Net (Regression) pycaret.regression
    • K-Means pycaret.clustering
    • Density-Based Spatial Clustering pycaret.clustering

  • Hyperparameter Tuning: New methods for hyperparameter tuning has been added in the tune_model function for pycaret.classification and pycaret.regression. New parameter search_library and search_algorithm in the tune_model function is added. search_library can be scikit-learn, scikit-optimize, tune-sklearn, and optuna. The search_algorithm param can take the following values based on its search_library:

    • scikit-learn: random grid
    • scikit-optimize: bayesian
    • tune-sklearn: random grid bayesian hyperopt bohb
    • optuna: random tpe

    Except for scikit-learn, all the other search libraries are not hard dependencies of pycaret and must be installed separately.

  • Early Stopping: Early stopping now supported for hyperparameter tuning. A new parameter early_stopping is added in the tune_model function for pycaret.classification and pycaret.regression. It is ignored when search_library is scikit-learn, or if the estimator doesn't have a 'partial_fit' attribute. It can be either an object accepted by the search library or one of the following:

    • asha for Asynchronous Successive Halving Algorithm
    • hyperband for Hyperband
    • median for median stopping rule
    • When False or None, early stopping will not be used.

  • Iterative Imputation: Iterative imputation type for numeric and categorical missing values is now implemented. New parameters imputation_type, iterative_imptutation_iters, categorical_iterative_imputer, and numeric_iterative_imputer added in the setup function. Read the blog post for more details: https://www.linkedin.com/pulse/iterative-imputation-pycaret-22-antoni-baum/?trackingId=Shg1zF%2F%2FR5BE7XFpzfTHkA%3D%3D

  • New Plots: Following new plots have been added:

    • lift pycaret.classification
    • gain pycaret.classification
    • tree pycaret.classification pycaret.regression
    • feature_all pycaret.classification pycaret.regression

  • CatBoost Compatibility: CatBoostClassifier and CatBoostRegressor is now compatible with plot_model. It requires catboost>=0.23.2.

  • Log Plots in MLFlow Server: You can now log any plot in the MLFlow tracking server that is available in the plot_model function. To log specific plots, pass a list containing plot IDs in the log_plots parameter. Check the documentation of the plot_model to see all available plots.

  • Data Split Stratification: A new parameter data_split_stratify is added in the setup function of pycaret.classification and pycaret.regression. It controls stratification during train_test_split. When set to True, will stratify by target column. To stratify on any other columns, pass a list of column names.

  • Fold Strategy: A new parameter fold_strategy is added in the setup function for pycaret.classification and pycaret.regression. By default, it is 'stratifiedkfold' for pycaret.classification and 'kfold' for pycaret.regression. Possible values are:

    • kfold for KFold CV;
    • stratifiedkfold for Stratified KFold CV;
    • groupkfold for Group KFold CV;
    • timeseries for TimeSeriesSplit CV; or
    • a custom CV generator object compatible with scikit-learn.

  • Global Fold Parameter: A new parameter fold has been added in the setup function for pycaret.classification and pycaret.regression. It controls the number of folds to be used in cross validation. This is a global setting that can be over-written at function level by using fold parameter within each function. Ignored when fold_strategy is a custom object.

  • Fold Groups: Optional Group labels when fold_strategy is groupkfold. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing the group label.

  • Transformation Pipeline: All transformations are now applied after train_test_split.

  • Data Type Handling: All data types handling internally has been changed from int64 and float64 to int32 and float32 respectively in order to improve memory usage and performance, as well as for better compatibility with GPU-based algorithms.

  • AutoML Behavior Change: automl function in pycaret.classification and pycaret.regression is no more re-fitting the model on the entire dataset. As such, if the model needs to be fitted on the entire dataset including the holdout set, finalize_model must be explicitly used.

  • Default Tuning Grid: Default hyperparameter tuning grid for RandomForest, XGBoost, CatBoost, and LightGBM has been amended to remove extreme values for max_depth and other training intense parameters to speed up the tuning process.

  • Random Forest Default Values: Default value of n_estimators for RandomForestClassifier and RandomForestRegressor has been changed from 10 to 100 to make it consistent with the default behavior of scikit-learn.

  • AUC for Multiclass Classification: AUC for Multiclass target is now available in the metric evaluation.

  • Google Colab Display: All output printed on screen (information grid, score grids) is now format compatible with Google Colab resulting in semantic improvements.

  • Sampling Parameter Removed: sampling parameter is now removed from the setup function of pycaret.classification and pycaret.regression.

  • Type Hinting: In order to make both the usage and development easier, type hints have been added to all updated pycaret functions, in accordance with best practices. Users can leverage those by using an IDE with support for type hints.

  • Documentation: All Modules documentation on the website is now retired. Updated documentation is available here: https://pycaret.readthedocs.io/en/latest/

Function Level Changes

New Functions Introduced in PyCaret 2.2

  • get_metrics: Returns table of available metrics used for CV. pycaret.classification pycaret.regression pycaret.clustering

  • add_metric: Adds a custom metric for model evaluation. pycaret.classification pycaret.regression pycaret.clustering

  • remove_metric: Remove custom metrics. pycaret.classification pycaret.regression pycaret.clustering

  • save_config: save all global variables to a pickle file, allowing to later resume without rerunning the setup function. pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly

  • load_config: Load global variables from pickle file into Python environment. pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly

setup

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly

Following new parameters have been added:

  • test_data: pandas.DataFrame, default = None If not None, test_data is used as a hold-out set, and the train_size parameter is ignored. test_data must be labeled and the shape of data and test_data must match.

  • preprocess: bool, default = True When set to False, no transformations are applied except for train_test_split and custom transformations passed in custom_pipeline param. Data must be ready for modeling (no missing values, no dates, categorical data encoding) when preprocess is set to False.

  • imputation_type: str, default = 'simple' The type of imputation to use. Can be either 'simple' or 'iterative'.

  • iterative_imputation_iters: int, default = 5 The number of iterations. Ignored when imputation_type is not 'iterative'.

  • categorical_iterative_imputer: str, default = 'lightgbm' Estimator for iterative imputation of missing values in categorical features. Ignored when imputation_type is not 'iterative'.

  • numeric_iterative_imputer: str, default = 'lightgbm' Estimator for iterative imputation of missing values in numeric features. Ignored when imputation_type is set to 'simple'.

  • data_split_stratify: bool or list, default = False Controls stratification during 'train_test_split'. When set to True, will stratify by target column. To stratify on any other columns, pass a list of column names. Ignored when data_split_shuffle is False.

  • fold_strategy: str or sklearn CV generator object, default = 'stratifiedkfold' / 'kfold' Choice of cross validation strategy. Possible values are:

    • 'kfold'
    • 'stratifiedkfold'
    • 'groupkfold'
    • 'timeseries'
    • a custom CV generator object compatible with scikit-learn.

  • fold: int, default = 10 The number of folds to be used in cross-validation. Must be at least 2. This is a global setting that can be over-written at the function level by using the fold parameter. Ignored when fold_strategy is a custom object.

  • fold_shuffle: bool, default = False Controls the shuffle parameter of CV. Only applicable when fold_strategy is 'kfold' or 'stratifiedkfold'. Ignored when fold_strategy is a custom object.

  • fold_groups: str or array-like, with shape (n_samples,), default = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

  • use_gpu: str or bool, default = False When set to 'force', will try to use GPU with all algorithms that support it, and raise exceptions if they are unavailable. When set to True, will use GPU with algorithms that support it, and fall back to CPU if they are unavailable. When False, all algorithms are trained using CPU only.

  • custom_pipeline: transformer or list of transformers or tuple, default = None* When passed, will append the custom transformers in the preprocessing pipeline and are applied on each CV fold separately and on the final fit. All the custom transformations are applied after 'train_test_split' and before pycaret's internal transformations.

compare_models

pycaret.classification pycaret.regression

Following new parameters have been added:

  • cross_validation: bool = True When set to False, metrics are evaluated on holdout set. fold param is ignored when cross_validation is set to False.

  • errors: str = "ignore" When set to 'ignore', will skip the model with exceptions and continue. If 'raise', will stop the function when exceptions are raised.

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

create_model

pycaret.classification pycaret.regression

Following new parameters have been added:

  • cross_validation: bool = True When set to False, metrics are evaluated on holdout set. fold param is ignored when cross_validation is set to False.

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

Following parameters have been removed:

  • ensemble - Deprecated - use ensemble_model function directly.
  • method - Deprecated - use ensemble_model function directly.
  • system - Moved to private API.

tune_model

pycaret.classification pycaret.regression

Following new parameters have been added:

  • search_library: str, default = 'scikit-learn' The search library used for tuning hyperparameters. Possible values:

    'scikit-learn' - default, requires no further installation https://github.com/scikit-learn/scikit-learn

    'scikit-optimize' - pip install scikit-optimize https://scikit-optimize.github.io/stable/

    'tune-sklearn' - pip install tune-sklearn ray[tune] https://github.com/ray-project/tune-sklearn

    'optuna' - pip install optuna https://optuna.org/

  • search_algorithm: str, default = None The search algorithm depends on the search_library parameter. Some search algorithms require additional libraries to be installed. When None, will use the search library-specific default algorithm.

    scikit-learn possible values: - random (default) - grid

    scikit-optimize possible values: - bayesian (default)

    tune-sklearn possible values: - random (default) - grid - bayesian pip install scikit-optimize - hyperopt pip install hyperopt - bohb pip install hpbandster ConfigSpace

    optuna possible values: - tpe (default) - random

  • early_stopping: bool or str or object, default = False Use early stopping to stop fitting to a hyperparameter configuration if it performs poorly. Ignored when search_library is scikit-learn, or if the estimator does not have 'partial_fit' attribute. If False or None, early stopping will not be used. Can be either an object accepted by the search library or one of the following:

    • 'asha' for Asynchronous Successive Halving Algorithm
    • 'hyperband' for Hyperband
    • 'median' for Median Stopping Rule
    • If False or None, early stopping will not be used.

  • early_stopping_max_iters: int, default = 10 The maximum number of epochs to run for each sampled configuration. Ignored if early_stopping is False or None.

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

  • return_tuner: bool, default = False When set to True, will return a tuple of (model, tuner_object).

  • tuner_verbose: bool or in, default = True If True or above 0, will print messages from the tuner. Higher values print more messages. Ignored when verbose param is False.

ensemble_model

pycaret.classification pycaret.regression

Following new parameters have been added:

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

blend_models

pycaret.classification pycaret.regression

Following new parameters have been added:

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

  • weights: list, default = None Sequence of weights (float or int) to weight the occurrences of predicted class labels (hard voting) or class probabilities before averaging (soft voting). Uses uniform weights when None.

  • The default value for the method parameter has been changed from hard to auto.

stack_models

pycaret.classification pycaret.regression

Following new parameters have been added:

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

calibrate_model

pycaret.classification

Following new parameters have been added:

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

plot_model

pycaret.classification pycaret.regression

Following new parameters have been added:

  • fold: int or scikit-learn compatible CV generator, default = None Controls cross-validation. If None, the CV generator in the fold_strategy parameter of the setup function is used. When an integer is passed, it is interpreted as the 'n_splits' parameter of the CV generator in the setup function.

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

evaluate_model

pycaret.classification pycaret.regression

Following new parameters have been added:

  • fold: int or scikit-learn compatible CV generator, default = None Controls cross-validation. If None, the CV generator in the fold_strategy parameter of the setup function is used. When an integer is passed, it is interpreted as the 'n_splits' parameter of the CV generator in the setup function.

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

finalize_model

pycaret.classification pycaret.regression

Following new parameters have been added:

  • fit_kwargs: Optional[dict] = None Dictionary of arguments passed to the fit method of the model.

  • groups: Optional[Union[str, Any]] = None Optional group labels when 'GroupKFold' is used for the cross-validation. It takes an array with shape (n_samples, ) where n_samples is the number of rows in the training dataset. When a string is passed, it is interpreted as the column name in the dataset containing group labels.

  • model_only: bool, default = True When set to False, only the model object is re-trained and all the transformations in Pipeline are ignored.

models

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly

Following new parameters have been added:

  • internal: bool, default = False When True, will return extra columns and rows used internally.

  • raise_errors: bool, default = True When False, will suppress all exceptions, ignoring models that couldn't be created.


Release: PyCaret 2.1.2 | Release Date: August 31, 2020 (BUG FIX)

  • Post-release 2.1 a bug has been reported preventing predict_model function to work in regression module in a new notebook session, when transform_target was set to False during model training. This issue has been fixed in PyCaret release 2.1.2. To learn more about the issue: pycaret#525


Release: PyCaret 2.1.1 | Release Date: August 30, 2020 (BUG FIX)

  • Post-release 2.1 a bug has been identified in MLFlow back-end. The error is only caused when log_experiment in the setup function is set to True and is applicable to all the modules. The cause of the error has been identified and an issue is opened with MLFlow. The error is caused by infer_signature function in mlflow.sklearn.log_model and is only raised when there are missing values in the dataset. This issue has been fixed in PyCaret release 2.1.1 by skipping the signature in cases where MLFlow raises exception.


Release: PyCaret 2.1 | Release Date: August 28, 2020

Summary of Changes

  • Model Deployment Model deployment support for gcp and azure has been added in deploy_model function for all modules. See documentation for details.
  • Compare Models Budget Time new parameter budget_time added in compare_models function. To set the upper limit on compare_models training time, budget_time parameter can be used.
  • Feature Selection New feature selection method boruta has been added for feature selection. By default, feature_selection_method parameter in the setup function is set to classic but can be set to boruta for feature selection using boruta algorithm. This change is applicable for pycaret.classification and pycaret.regression.
  • Numeric Imputation New method zero has been added in the numeric_imputation in the setup function. When method is set to zero, missing values are replaced with constant 0. Default behavior of numeric_imputation is unchanged.
  • Plot Model New parameter scale has been added in plot_model for all modules to enable high quality images for research publications.
  • User Defined Loss Function You can now pass custom_scorer for optimizing user defined loss function in tune_model for pycaret.classification and pycaret.regression. You must use make_scorer from sklearn to create custom loss function that can be passed into custom_scorer for the tune_model function.
  • Change in Pipeline Behavior When using save_model the model object is appended into Pipeline, as such the behavior of Pipeline and predict_model is now changed. Instead of saving a list, save_model now saves Pipeline object where trained model is on last position. The user functionality on front-end for predict_model remains same.
  • Compare Models parameter blacklist and whitelist is now renamed to exclude and include with no change in functionality.
  • Predict Model Labels The Label column returned by predict_model function in pycaret.classification now returns the original label instead of encoded value. This change is made to make output from predict_model more human-readable. A new parameter encoded_labels is added, which is False by default. When set to True, it will return encoded labels.
  • Model Logging Model persistence in the backend when log_experiment is set to True is now changed. Instead of using internal save_model functionality, it now adopts to mlflow.sklearn.save_model to allow the use of Model Registry and MLFlow native deployment functionalities.
  • CatBoost Compatibility CatBoostClassifier is now compatible with blend_models in pycaret.classification. As such blend_models without any estimator_list will now result in blending total of 15 estimators including CatBoostClassifier.
  • Stack Models stack_models in pycaret.classification and pycaret.regression now adopts to StackingClassifier() and StackingRegressor from sklearn. As such the stack_models function now returns sklearn object instead of custom list in previous versions.
  • Create Stacknet create_stacknet in pycaret.classification and pycaret.regression is now removed.
  • Tune Model tune_model in pycaret.classification and pycaret.regression now inherits params from the input estimator. As such if you have trained xgboost, lightgbm or catboost on gpu will not inherits training method from estimator.
  • Interpret Model **kwargs argument now added in interpret_model.
  • Pandas Categorical Type All modules are now compatible with pandas.Categorical object. Internally they are converted into object and are treated as the same way as object or bool is treated.
  • use_gpu A new parameter added in the setup function for pycaret.classification and pycaret.regression. In 2.1 it was added to prepare for the backend work required to make this change in future releases. As such using use_gpu param in 2.1 has no impact.
  • Unit Tests Unit testing enhanced. Continious improvement in progress https://github.com/pycaret/pycaret/tree/master/pycaret/tests
  • Automated Documentation Added Automated documentation now added. Documentation on Website will only update for major releases 0.X. For all minor monthly releases, documentation will be available on: https://pycaret.readthedocs.io/en/latest/
  • Introduction of GitHub Actions CI/CD build testing is now moved from travis-ci to github-actions. pycaret-nightly is now being published every 24 hours automatically.
  • Tutorials All tutorials are now updated using pycaret==2.0. https://github.com/pycaret/pycaret/tree/master/tutorials
  • Resources New resources added under /pycaret/resources/ https://github.com/pycaret/pycaret/tree/master/resources
  • Example Notebook Many example notebooks added under /pycaret/examples/ https://github.com/pycaret/pycaret/tree/master/examples


Release: PyCaret 2.0 | Release Date: July 31, 2020

Summary of Changes

  • Experiment Logging MLFlow logging backend added. New parameters log_experiment experiment_name log_profile log_data added in setup. Available in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp
  • Save / Load Experiment save_experiment and load_experiment function from pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp is removed in PyCaret 2.0
  • System Logging System log files now generated when setup is executed. logs.log file is saved in current working directory. Function get_system_logs can be used to access log file in notebook.
  • Command Line Support When using PyCaret 2.0 outside of Notebook, html parameter in setup must be set to False.
  • Imbalance Dataset fix_imbalance and fix_imbalance_method parameter added in setup for pycaret.classification. When set to True, SMOTE is applied by default to create synthetic datapoints for minority class. To change the method pass any class from imblearn that supports fit_resample method in fix_imbalance_method parameter.
  • Save Plot save parameter added in plot_model. When set to True, it saves the plot as png or html in current working directory.
  • kwargs kwargs** added in create_model for pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly
  • choose_better choose_better and optimize parameter added in tune_model ensemble_model blend_models stack_models create_stacknet in pycaret.classification and pycaret.regression. Read the details below to learn more about thi added in create_model for pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly
  • Training Time TT (Sec) added in compare_models function for pycaret.classification and pycaret.regression
  • New Metric: MCC MCC metric added in score grid for pycaret.classification
  • NEW FUNCTION: automl() New function automl added in pycaret.classification pycaret.regression
  • NEW FUNCTION: pull() New function pull added in pycaret.classification pycaret.regression
  • NEW FUNCTION: models() New function models added in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp
  • NEW FUNCTION: get_logs() New function get_logs added in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp
  • NEW FUNCTION: get_config() New function get_config added in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp
  • NEW FUNCTION: set_config() New function set_config added in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp
  • NEW FUNCTION: get_system_logs New function get_logs added in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp
  • CHANGE IN BEHAVIOR: compare_models compare_models now returns top_n models defined by n_select parameter, by default set to 1.
  • CHANGE IN BEHAVIOR: tune_model tune_model function in pycaret.classification and pycaret.regression now requires trained model object to be passed as estimator instead of string abbreviation / ID.
  • REMOVED DEPENDENCIES awscli and shap removed from requirements.txt. To use interpret_model function in pycaret.classification pycaret.regression and deploy_model function in pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly, these libraries will have to be installed separately.

setup

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • remove_perfect_collinearity parameter added in setup(). Default set to False.
    When set to True, perfect collinearity (features with correlation = 1) is removed from the dataset, When two features are 100% correlated, one of it is randomly dropped from the dataset.

  • fix_imbalance parameter added in setup(). Default set to False.
    When dataset has unequal distribution of target class it can be fixed using fix_imbalance parameter. When set to True, SMOTE (Synthetic Minority Over-sampling Technique) is applied by default to create synthetic datapoints for minority class.

  • fix_imbalance_method parameter added in setup(). Default set to None.
    When fix_imbalance is set to True and fix_imbalance_method is None, 'smote' is applied by default to oversample minority class during cross validation. This parameter accepts any module from 'imblearn' that supports 'fit_resample' method.

  • data_split_shuffle parameter added in setup(). Default set to True.
    If set to False, prevents shuffling of rows when splitting data.

  • folds_shuffle parameter added in setup(). Default set to False.
    If set to False, prevents shuffling of rows when using cross validation.

  • n_jobs parameter added in setup(). Default set to -1.
    The number of jobs to run in parallel (for functions that supports parallel processing) -1 means using all processors. To run all functions on single processor set n_jobs to None.

  • html parameter added in setup(). Default set to True.
    If set to False, prevents runtime display of monitor. This must be set to False when using environment that doesnt support HTML.

  • log_experiment parameter added in setup(). Default set to False.
    When set to True, all metrics and parameters are logged on MLFlow server.

  • experiment_name parameter added in setup(). Default set to None.
    Name of experiment for logging. When set to None, 'clf' is by default used as alias for the experiment name.

  • log_plots parameter added in setup(). Default set to False.
    When set to True, specific plots are logged in MLflow as a png file.

  • log_profile parameter added in setup(). Default set to False.
    When set to True, data profile is also logged on MLflow as a html file.

  • log_data parameter added in setup(). Default set to False.
    When set to True, train and test dataset are logged as csv.

  • verbose parameter added in setup(). Default set to True.
    Information grid is not printed when verbose is set to False.

compare_models

pycaret.classification pycaret.regression

  • whitelist parameter added in compare_models. Default set to None.
    In order to run only certain models for the comparison, the model ID's can be passed as a list of strings in whitelist param.

  • n_select parameter added in compare_models. Default set to 1.
    Number of top_n models to return. use negative argument for bottom selection. For example, n_select = -3 means bottom 3 models.

  • verbose parameter added in compare_models. Default set to True.
    Score grid is not printed when verbose is set to False.

create_model

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly

  • cross_validation parameter added in create_model. Default set to True.
    When cross_validation set to False fold parameter is ignored and model is trained on entire training dataset. No metric evaluation is returned. Only applicable in pycaret.classification and pycaret.regression

  • system parameter added in create_model. Default set to True.
    Must remain True all times. Only to be changed by internal functions.

  • ground_truth parameter added in create_model. Default set to None.
    When ground_truth is provided, Homogeneity Score, Rand Index, and Completeness Score is evaluated and printer along with other metrics. This is only available in pycaret.clustering

  • kwargs parameter added in create_model.
    Additional keyword arguments to pass to the estimator.

tune_model

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • custom_grid parameter added in tune_model. Default set to None.
    To use custom hyperparameters for tuning pass a dictionary with parameter name and values to be iterated. When set to None it uses pre-defined tuning grid. For pycaret.clustering pycaret.anomaly pycaret.nlp, custom_grid param must be a list of values to iterate over.

  • choose_better parameter added in tune_model. Default set to False.
    When set to set to True, base estimator is returned when the performance doesn't improve by tune_model. This gurantees the returned object would perform atleast equivalent to base estimator created using create_model or model returned by compare_models.

ensemble_model

pycaret.classification pycaret.regression

  • choose_better parameter added in ensemble_model. Default set to False.
    When set to set to True, base estimator is returned when the performance doesn't improve by tune_model. This gurantees the returned object would perform atleast equivalent to base estimator created using create_model or model returned by compare_models.

  • optimize parameter added in ensemble_model. Default set to Accuracy for pycaret.classification and R2 for pycaret.regression.
    Only used when choose_better is set to True. optimize parameter is used to compare emsembled model with base estimator. Values accepted in optimize parameter for pycaret.classification are 'Accuracy', 'AUC', 'Recall', 'Precision', 'F1', 'Kappa', 'MCC' and for pycaret.regression are 'MAE', 'MSE', 'RMSE' 'R2', 'RMSLE' and 'MAPE'.

blend_models

pycaret.classification pycaret.regression

  • choose_better parameter added in blend_models. Default set to False.
    When set to set to True, base estimator is returned when the performance doesn't improve by tune_model. This gurantees the returned object would perform atleast equivalent to base estimator created using create_model or model returned by compare_models.

  • optimize parameter added in blend_models. Default set to Accuracy for pycaret.classification and R2 for pycaret.regression.
    Only used when choose_better is set to True. optimize parameter is used to compare emsembled model with base estimator. Values accepted in optimize parameter for pycaret.classification are 'Accuracy', 'AUC', 'Recall', 'Precision', 'F1', 'Kappa', 'MCC' and for pycaret.regression are 'MAE', 'MSE', 'RMSE' 'R2', 'RMSLE' and 'MAPE'.

stack_models

pycaret.classification pycaret.regression

  • choose_better parameter added in stack_models. Default set to False.
    When set to set to True, base estimator is returned when the performance doesn't improve by tune_model. This gurantees the returned object would perform atleast equivalent to base estimator created using create_model or model returned by compare_models.

  • optimize parameter added in stack_models. Default set to Accuracy for pycaret.classification and R2 for pycaret.regression.
    Only used when choose_better is set to True. optimize parameter is used to compare emsembled model with base estimator. Values accepted in optimize parameter for pycaret.classification are 'Accuracy', 'AUC', 'Recall', 'Precision', 'F1', 'Kappa', 'MCC' and for pycaret.regression are 'MAE', 'MSE', 'RMSE' 'R2', 'RMSLE' and 'MAPE'.

create_stacknet

pycaret.classification pycaret.regression

  • choose_better parameter added in create_stacknet. Default set to False.
    When set to set to True, base estimator is returned when the performance doesn't improve by tune_model. This gurantees the returned object would perform atleast equivalent to base estimator created using create_model or model returned by compare_models.

  • optimize parameter added in create_stacknet. Default set to Accuracy for pycaret.classification and R2 for pycaret.regression.
    Only used when choose_better is set to True. optimize parameter is used to compare emsembled model with base estimator. Values accepted in optimize parameter for pycaret.classification are 'Accuracy', 'AUC', 'Recall', 'Precision', 'F1', 'Kappa', 'MCC' and for pycaret.regression are 'MAE', 'MSE', 'RMSE' 'R2', 'RMSLE' and 'MAPE'.

predict_model

pycaret.classification pycaret.regression

  • verbose parameter added in predict_model. Default set to True.
    Holdout score grid is not printed when verbose is set to False.

plot_model

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • save parameter added in plot_model. Default set to False.
    When set to True, Plot is saved as a 'png' file in current working directory.

  • verbose parameter added in plot_model. Default set to True.
    Progress bar not shown when verbose set to False.

  • system parameter added in plot_model. Default set to True.
    Must remain True all times. Only to be changed by internal functions.

NEW FUNCTION: automl

pycaret.classification pycaret.regression

  • This function returns the best model out of all models created in current active environment based on metric defined in optimize parameter.

Parameters:

  • optimize string, default = 'Accuracy' for pycaret.classification and 'R2' for pycaret.regression
    Other values you can pass in optimize param are 'AUC', 'Recall', 'Precision', 'F1', 'Kappa', and 'MCC' for pycaret.classification and 'MAE', 'MSE', 'RMSE', 'R2', 'RMSLE', and 'MAPE' for pycaret.regression

  • use_holdout bool, default = False
    When set to True, metrics are evaluated on holdout set instead of CV.

NEW FUNCTION: pull

pycaret.classification pycaret.regression

  • This function returns the last printed score grid as pandas dataframe.

NEW FUNCTION: models

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • This function Returns the table of models available in model library.

Parameters:

  • type string, default = None
    linear : filters and only return linear models
    tree : filters and only return tree based models
    ensemble : filters and only return ensemble models

type parameter only available in pycaret.classification and pycaret.regression

NEW FUNCTION: get_logs

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • This function returns a table with experiment logs consisting run details, parameter, metrics and tags.

Parameters:

  • experiment_name string, default = None
    When set to None current active run is used.

  • save bool, default = False
    When set to True, csv file is saved in current directory.

NEW FUNCTION: get_config

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • This function is used to access global environment variables. Check docstring for the list of global var accessible.

NEW FUNCTION: set_config

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • This function is used to reset global environment variables. Check docstring for the list of global var accessible.

NEW FUNCTION: get_system_logs

pycaret.classification pycaret.regression pycaret.clustering pycaret.anomaly pycaret.nlp

  • This function is reads and print 'logs.log' file from current active directory. logs.log is generated from setup is initialized in any module.