Scikit-Learn Model Selection Patterns

In [1]:

    

import pandas as pd
from sklearn.datasets import load_iris

X, y = load_iris(return_X_y=True)


    

In [122]:

    

from sklearn.pipeline import Pipeline
from sklearn.model_selection import GridSearchCV
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler

# Make the model
pipeline = Pipeline([
    ('scale', None), 
    ('predict', None)
])

# Set up some parameters
params = [
    {
    'scale':[None, StandardScaler()],
    'predict': [LogisticRegression()],
    'predict__C': [0.1, 1, 10, 100, 1000]
    },
    {
    'scale':[None, StandardScaler()],
    'predict': [SVC()],
    'predict__C': [0.1, 1, 10, 100, 1000]
}
]

gs = GridSearchCV(pipeline, params, scoring='accuracy', cv=StratifiedKFold(), return_train_score=True)
gs.fit(X, y)

res = pd.DataFrame(gs.cv_results_)
pd.options.display.max_colwidth=100
print(gs.best_estimator_)
res[['param_scale', 'param_predict', 'mean_test_score', 'std_test_score']]


    

    




Pipeline(memory=None,
     steps=[('scale', StandardScaler(copy=True, with_mean=True, with_std=True)), ('predict', SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,
  decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',
  max_iter=-1, probability=False, random_state=None, shrinking=True,
  tol=0.001, verbose=False))])






    
Out[122]:







  

    

      

      
param_scale
      
param_predict
      
mean_test_score
      
std_test_score
    
  
  

    

      
0
      
None
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.813333
      
0.037924
    
    

      
1
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.833333
      
0.053703
    
    

      
2
      
None
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.946667
      
0.018041
    
    

      
3
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.900000
      
0.016420
    
    

      
4
      
None
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.966667
      
0.033456
    
    

      
5
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.926667
      
0.017795
    
    

      
6
      
None
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.973333
      
0.037154
    
    

      
7
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.960000
      
0.027588
    
    

      
8
      
None
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.966667
      
0.046442
    
    

      
9
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
LogisticRegression(C=1000, class_weight=None, dual=False, fit_intercept=True,\n          interce...
      
0.960000
      
0.027588
    
    

      
10
      
None
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.953333
      
0.023989
    
    

      
11
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.920000
      
0.030870
    
    

      
12
      
None
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.973333
      
0.009021
    
    

      
13
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.953333
      
0.036876
    
    

      
14
      
None
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.980000
      
0.015925
    
    

      
15
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.986667
      
0.018577
    
    

      
16
      
None
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.960000
      
0.042507
    
    

      
17
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.966667
      
0.024254
    
    

      
18
      
None
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.953333
      
0.036876
    
    

      
19
      
StandardScaler(copy=True, with_mean=True, with_std=True)
      
SVC(C=10, cache_size=200, class_weight=None, coef0=0.0,\n  decision_function_shape='ovr', degree...
      
0.960000
      
0.015435
    
  

In [ ]: