PyCaret 2.0x New Features¶

In [ ]:

#pip install pycaret-nightly

In [1]:

from pycaret.utils import version
version()

pycaret-nightly-0.26

In [2]:

from pycaret.datasets import get_data
data = get_data('juice')

	Id	Purchase	WeekofPurchase	StoreID	PriceCH	PriceMM	DiscCH	DiscMM	SpecialMM	LoyalCH	SalePriceMM	SalePriceCH	PriceDiff	Store7	PctDiscMM	PctDiscCH	ListPriceDiff	STORE
0	1	CH	237	1	1.75	1.99	0.00	0.0	0	0.500000	1.99	1.75	0.24	No	0.000000	0.000000	0.24	1
1	2	CH	239	1	1.75	1.99	0.00	0.3	1	0.600000	1.69	1.75	-0.06	No	0.150754	0.000000	0.24	1
2	3	CH	245	1	1.86	2.09	0.17	0.0	0	0.680000	2.09	1.69	0.40	No	0.000000	0.091398	0.23	1
3	4	MM	227	1	1.69	1.69	0.00	0.0	0	0.400000	1.69	1.69	0.00	No	0.000000	0.000000	0.00	1
4	5	CH	228	7	1.69	1.69	0.00	0.0	0	0.956535	1.69	1.69	0.00	Yes	0.000000	0.000000	0.00	0

In [3]:

from pycaret.classification import *
clf1 = setup(data, target = 'Purchase', session_id=7267, log_experiment=True, experiment_name='pycaret2-juice')

Setup Succesfully Completed!

	Description	Value
0	session_id	7267
1	Target Type	Binary
2	Label Encoded	CH: 0, MM: 1
3	Original Data	(1070, 19)
4	Missing Values	False
5	Numeric Features	13
6	Categorical Features	5
7	Ordinal Features	False
8	High Cardinality Features	False
9	High Cardinality Method	None
10	Sampled Data	(1070, 19)
11	Transformed Train Set	(748, 16)
12	Transformed Test Set	(322, 16)
13	Numeric Imputer	mean
14	Categorical Imputer	constant
15	Normalize	False
16	Normalize Method	None
17	Transformation	False
18	Transformation Method	None
19	PCA	False
20	PCA Method	None
21	PCA Components	None
22	Ignore Low Variance	False
23	Combine Rare Levels	False
24	Rare Level Threshold	None
25	Numeric Binning	False
26	Remove Outliers	False
27	Outliers Threshold	None
28	Remove Multicollinearity	False
29	Multicollinearity Threshold	None
30	Clustering	False
31	Clustering Iteration	None
32	Polynomial Features	False
33	Polynomial Degree	None
34	Trignometry Features	False
35	Polynomial Threshold	None
36	Group Features	False
37	Feature Selection	False
38	Features Selection Threshold	None
39	Feature Interaction	False
40	Feature Ratio	False
41	Interaction Threshold	None
42	Fix Imbalance	False
43	Fix Imbalance Method	SMOTE

1. Behavior of compare_models¶

1.1 Default¶

In [4]:

best_model = compare_models()

	Model	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC	TT (Sec)
0	Linear Discriminant Analysis	0.8383	0.9083	0.7978	0.7933	0.7934	0.6607	0.6632	0.0161
1	Ridge Classifier	0.8356	0.0000	0.7979	0.7873	0.7907	0.6555	0.6578	0.0295
2	Ada Boost Classifier	0.8276	0.8913	0.7637	0.7922	0.7760	0.6360	0.6380	0.2087
3	CatBoost Classifier	0.8223	0.9026	0.7502	0.7884	0.7672	0.6238	0.6261	5.0027
4	Logistic Regression	0.8222	0.9071	0.7430	0.7927	0.7643	0.6221	0.6255	0.1138
5	Gradient Boosting Classifier	0.8222	0.8967	0.7637	0.7777	0.7697	0.6251	0.6261	0.2643
6	Light Gradient Boosting Machine	0.8089	0.8865	0.7469	0.7609	0.7529	0.5973	0.5982	0.3430
7	Random Forest Classifier	0.8035	0.8672	0.7163	0.7672	0.7393	0.5824	0.5847	0.1235
8	Extreme Gradient Boosting	0.8035	0.8776	0.7464	0.7533	0.7480	0.5872	0.5892	0.2332
9	Extra Trees Classifier	0.7981	0.8468	0.7159	0.7558	0.7343	0.5718	0.5734	0.3468
10	Naive Bayes	0.7859	0.8488	0.8185	0.6920	0.7493	0.5649	0.5718	0.0068
11	Decision Tree Classifier	0.7566	0.7429	0.6780	0.6942	0.6847	0.4869	0.4881	0.0181
12	K Neighbors Classifier	0.7393	0.7890	0.6098	0.6910	0.6447	0.4409	0.4454	0.0138
13	Quadratic Discriminant Analysis	0.7286	0.7917	0.6844	0.6699	0.6460	0.4340	0.4504	0.0121
14	SVM - Linear Kernel	0.5643	0.0000	0.2000	0.0773	0.1115	0.0000	0.0000	0.0319

In [5]:

best_model

Out[5]:

LinearDiscriminantAnalysis(n_components=None, priors=None, shrinkage=None,
                           solver='svd', store_covariance=False, tol=0.0001)

1.2 n_select parameter¶

In [ ]:

top5 = compare_models(n_select=5)

In [7]:

Out[7]:

LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
               importance_type='split', learning_rate=0.1, max_depth=-1,
               min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
               n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
               random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
               subsample=1.0, subsample_for_bin=200000, subsample_freq=0)

1.3 Whitelist parameter¶

In [6]:

w = compare_models(whitelist=['dt','rf','xgboost','lightgbm'])

	Model	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC	TT (Sec)
0	Light Gradient Boosting Machine	0.8089	0.8865	0.7469	0.7609	0.7529	0.5973	0.5982	0.3285
1	Random Forest Classifier	0.8035	0.8672	0.7163	0.7672	0.7393	0.5824	0.5847	0.1270
2	Extreme Gradient Boosting	0.8035	0.8776	0.7464	0.7533	0.7480	0.5872	0.5892	0.2405
3	Decision Tree Classifier	0.7566	0.7429	0.6780	0.6942	0.6847	0.4869	0.4881	0.0081

2. Behavior of tune_model¶

In [8]:

tuned_best_model = tune_model(best_model, optimize='Precision')

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0	0.8400	0.9445	0.7241	0.8400	0.7778	0.6538	0.6582
1	0.7867	0.8621	0.6552	0.7600	0.7037	0.5385	0.5421
2	0.8267	0.8808	0.7241	0.8077	0.7636	0.6274	0.6298
3	0.8533	0.9453	0.8621	0.7812	0.8197	0.6966	0.6990
4	0.8000	0.8808	0.6552	0.7917	0.7170	0.5645	0.5705
5	0.8133	0.9220	0.7241	0.7778	0.7500	0.6014	0.6023
6	0.8267	0.8830	0.7000	0.8400	0.7636	0.6286	0.6351
7	0.8133	0.9178	0.7000	0.8077	0.7500	0.6023	0.6062
8	0.8378	0.8720	0.7586	0.8148	0.7857	0.6555	0.6566
9	0.8108	0.9069	0.6897	0.8000	0.7407	0.5931	0.5971
Mean	0.8209	0.9015	0.7193	0.8021	0.7572	0.6162	0.6197
SD	0.0190	0.0285	0.0564	0.0245	0.0317	0.0440	0.0435

In [ ]:

tuned_best_model

3. Train custom models¶

In [9]:

from gplearn.genetic import SymbolicClassifier
sc = SymbolicClassifier()

In [10]:

sc = create_model(sc, fold=5)

IntProgress(value=0, description='Processing: ', max=9)



Initiated	. . . . . . . . . . . . . . . . . .	12:55:14
Status	. . . . . . . . . . . . . . . . . .	Creating Logs
ETC	. . . . . . . . . . . . . . . . . .	Almost Finished

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0	0.8267	0.9213	0.7241	0.8077	0.7636	0.6274	0.6298

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-10-ffd762ec6369> in <module>
----> 1 sc = create_model(sc, fold=5)

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\pycaret\classification.py in create_model(estimator, ensemble, method, fold, round, cross_validation, verbose, system, **kwargs)
   2422         if hasattr(model, 'predict_proba'):
   2423             logger.info("Fitting Model")
-> 2424             model.fit(Xtrain,ytrain)
   2425             logger.info("Evaluating Metrics")
   2426             pred_prob = model.predict_proba(Xtest)

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\gplearn\genetic.py in fit(self, X, y, sample_weight)
    469                                           seeds[starts[i]:starts[i + 1]],
    470                                           params)
--> 471                 for i in range(n_jobs))
    472 
    473             # Reduce, maintaining order across different n_jobs

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\parallel.py in __call__(self, iterable)
   1027             # remaining jobs.
   1028             self._iterating = False
-> 1029             if self.dispatch_one_batch(iterator):
   1030                 self._iterating = self._original_iterator is not None
   1031 

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\parallel.py in dispatch_one_batch(self, iterator)
    845                 return False
    846             else:
--> 847                 self._dispatch(tasks)
    848                 return True
    849 

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\parallel.py in _dispatch(self, batch)
    763         with self._lock:
    764             job_idx = len(self._jobs)
--> 765             job = self._backend.apply_async(batch, callback=cb)
    766             # A job can complete so quickly than its callback is
    767             # called before we get here, causing self._jobs to

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\_parallel_backends.py in apply_async(self, func, callback)
    204     def apply_async(self, func, callback=None):
    205         """Schedule a func to be run"""
--> 206         result = ImmediateResult(func)
    207         if callback:
    208             callback(result)

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\_parallel_backends.py in __init__(self, batch)
    568         # Don't delay the application, to avoid keeping the input
    569         # arguments in memory
--> 570         self.results = batch()
    571 
    572     def get(self):

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\parallel.py in __call__(self)
    251         with parallel_backend(self._backend, n_jobs=self._n_jobs):
    252             return [func(*args, **kwargs)
--> 253                     for func, args, kwargs in self.items]
    254 
    255     def __reduce__(self):

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\joblib\parallel.py in <listcomp>(.0)
    251         with parallel_backend(self._backend, n_jobs=self._n_jobs):
    252             return [func(*args, **kwargs)
--> 253                     for func, args, kwargs in self.items]
    254 
    255     def __reduce__(self):

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\gplearn\genetic.py in _parallel_evolve(n_programs, parents, X, y, sample_weight, seeds, params)
    144         oob_sample_weight[indices] = 0
    145 
--> 146         program.raw_fitness_ = program.raw_fitness(X, y, curr_sample_weight)
    147         if max_samples < n_samples:
    148             # Calculate OOB fitness

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\gplearn\_program.py in raw_fitness(self, X, y, sample_weight)
    460 
    461         """
--> 462         y_pred = self.execute(X)
    463         if self.transformer:
    464             y_pred = self.transformer(y_pred)

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\gplearn\_program.py in execute(self, X)
    378                              else X[:, t] if isinstance(t, int)
    379                              else t for t in apply_stack[-1][1:]]
--> 380                 intermediate_result = function(*terminals)
    381                 if len(apply_stack) != 1:
    382                     apply_stack.pop()

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\gplearn\functions.py in __call__(self, *args)
     44 
     45     def __call__(self, *args):
---> 46         return self.function(*args)
     47 
     48 

~\Anaconda3\envs\pycaret-nightly-env\lib\site-packages\gplearn\functions.py in _protected_division(x1, x2)
    125     """Closure of division (x1/x2) for zero denominator."""
    126     with np.errstate(divide='ignore', invalid='ignore'):
--> 127         return np.where(np.abs(x2) > 0.001, np.divide(x1, x2), 1.)
    128 
    129 

KeyboardInterrupt:

In [ ]:

predict_model(sc);

In [ ]:

plot_model(sc, plot = 'confusion_matrix')

In [ ]:

plot_model(sc, plot = 'auc')

In [ ]:

plot_model(sc, plot = 'threshold')

4. Customize parameters in create_model¶

In [11]:

lightgbm = create_model('lightgbm')

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0	0.7867	0.9078	0.6552	0.7600	0.7037	0.5385	0.5421
1	0.7867	0.8741	0.7586	0.7097	0.7333	0.5559	0.5567
2	0.7600	0.8501	0.7241	0.6774	0.7000	0.5004	0.5011
3	0.8800	0.9175	0.8621	0.8333	0.8475	0.7486	0.7489
4	0.8267	0.8703	0.7931	0.7667	0.7797	0.6369	0.6371
5	0.8400	0.9025	0.7931	0.7931	0.7931	0.6627	0.6627
6	0.7467	0.8541	0.6667	0.6897	0.6780	0.4693	0.4695
7	0.8133	0.9244	0.7333	0.7857	0.7586	0.6067	0.6077
8	0.8243	0.8506	0.7241	0.8077	0.7636	0.6245	0.6269
9	0.8243	0.9138	0.7586	0.7857	0.7719	0.6291	0.6294
Mean	0.8089	0.8865	0.7469	0.7609	0.7529	0.5973	0.5982
SD	0.0375	0.0282	0.0582	0.0494	0.0480	0.0784	0.0782

In [12]:

lightgbm

Out[12]:

LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
               importance_type='split', learning_rate=0.1, max_depth=-1,
               min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
               n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
               random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
               subsample=1.0, subsample_for_bin=200000, subsample_freq=0)

In [13]:

import numpy as np

lgbms = []

for i in np.arange(0.1,1,0.1):
    m = create_model('lightgbm', learning_rate=i, verbose=False)
    lgbms.append(m)

In [14]:

print(len(lgbms))

In [15]:

lgbms

Out[15]:

[LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.1, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.2, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.30000000000000004,
                max_depth=-1, min_child_samples=20, min_child_weight=0.001,
                min_split_gain=0.0, n_estimators=100, n_jobs=-1, num_leaves=31,
                objective=None, random_state=7267, reg_alpha=0.0, reg_lambda=0.0,
                silent=True, subsample=1.0, subsample_for_bin=200000,
                subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.4, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.5, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.6, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.7000000000000001,
                max_depth=-1, min_child_samples=20, min_child_weight=0.001,
                min_split_gain=0.0, n_estimators=100, n_jobs=-1, num_leaves=31,
                objective=None, random_state=7267, reg_alpha=0.0, reg_lambda=0.0,
                silent=True, subsample=1.0, subsample_for_bin=200000,
                subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.8, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0),
 LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,
                importance_type='split', learning_rate=0.9, max_depth=-1,
                min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,
                n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,
                random_state=7267, reg_alpha=0.0, reg_lambda=0.0, silent=True,
                subsample=1.0, subsample_for_bin=200000, subsample_freq=0)]

5. AutoML()¶

In [16]:

best_model_cv = automl()
print(best_model_cv)

LinearDiscriminantAnalysis(n_components=None, priors=None, shrinkage=None,
                           solver='svd', store_covariance=False, tol=0.0001)

In [ ]:

In [17]:

best_model_holdout = automl(use_holdout=True)
print(best_model_holdout)

LinearDiscriminantAnalysis(n_components=None, priors=None, shrinkage=None,
                           solver='svd', store_covariance=False, tol=0.0001)

6. MLFlow UI¶

In [18]:

!mlflow ui

^C

7. Log plots¶

In [19]:

from pycaret.classification import *
clf1 = setup(data, target = 'Purchase', session_id=7267, log_experiment=True, experiment_name='pycaret2-juice',
             log_plots = True)

Setup Succesfully Completed!

	Description	Value
0	session_id	7267
1	Target Type	Binary
2	Label Encoded	CH: 0, MM: 1
3	Original Data	(1070, 19)
4	Missing Values	False
5	Numeric Features	13
6	Categorical Features	5
7	Ordinal Features	False
8	High Cardinality Features	False
9	High Cardinality Method	None
10	Sampled Data	(1070, 19)
11	Transformed Train Set	(748, 16)
12	Transformed Test Set	(322, 16)
13	Numeric Imputer	mean
14	Categorical Imputer	constant
15	Normalize	False
16	Normalize Method	None
17	Transformation	False
18	Transformation Method	None
19	PCA	False
20	PCA Method	None
21	PCA Components	None
22	Ignore Low Variance	False
23	Combine Rare Levels	False
24	Rare Level Threshold	None
25	Numeric Binning	False
26	Remove Outliers	False
27	Outliers Threshold	None
28	Remove Multicollinearity	False
29	Multicollinearity Threshold	None
30	Clustering	False
31	Clustering Iteration	None
32	Polynomial Features	False
33	Polynomial Degree	None
34	Trignometry Features	False
35	Polynomial Threshold	None
36	Group Features	False
37	Feature Selection	False
38	Features Selection Threshold	None
39	Feature Interaction	False
40	Feature Ratio	False
41	Interaction Threshold	None
42	Fix Imbalance	False
43	Fix Imbalance Method	SMOTE

In [20]:

lr = create_model('lr')

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0	0.8267	0.9468	0.6552	0.8636	0.7451	0.6175	0.6310
1	0.7600	0.8778	0.7241	0.6774	0.7000	0.5004	0.5011
2	0.8267	0.8771	0.6897	0.8333	0.7547	0.6225	0.6292
3	0.8800	0.9400	0.8966	0.8125	0.8525	0.7517	0.7543
4	0.8133	0.8808	0.6552	0.8261	0.7308	0.5908	0.6001
5	0.8000	0.9138	0.7241	0.7500	0.7368	0.5756	0.5759
6	0.7867	0.8889	0.7000	0.7500	0.7241	0.5506	0.5514
7	0.8533	0.9296	0.8333	0.8065	0.8197	0.6961	0.6964
8	0.8243	0.8904	0.7586	0.7857	0.7719	0.6291	0.6294
9	0.8514	0.9261	0.7931	0.8214	0.8070	0.6862	0.6865
Mean	0.8222	0.9071	0.7430	0.7927	0.7643	0.6221	0.6255
SD	0.0330	0.0258	0.0744	0.0512	0.0456	0.0704	0.0703

<Figure size 576x396 with 0 Axes>

In [ ]:

8. Logs on excel¶

In [21]:

#useful for remote runs like Kaggle Kernel or GitHub actions
xl_logs = get_logs(save=True)

In [ ]:

9. Databricks integration¶

In [ ]:

#see example on Databricks