Python) Permutation Importance 다양하게 표현하는 방법

해당 파일에 함수를 정리하였습니다.

advanced_feature_importance.py

0.01MB

 

In [26]:

from IPython.core.display import display, HTML
display(HTML("<style>.container {width:90% !important;}</style>"))

 

 

/tmp/ipykernel_3209568/3510566465.py:1: DeprecationWarning: Importing display from IPython.core.display is deprecated since IPython 7.14, please import from IPython display
  from IPython.core.display import display, HTML

 

 

 

참고¶

해당 코드를 기반으로 정리한 코드입니다.

https://github.com/cerlymarco/MEDIUM_NoteBook/blob/master/Advanced_Perm_Importance/Advanced_Perm_Importance.ipynb

 

In [1]:

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from tqdm import tqdm

from sklearn.linear_model import *
from sklearn.datasets import fetch_openml, fetch_california_housing

 

In [2]:

from advanced_feature_importance import PermutationImportance 

 

Regression¶

 

In [3]:

### READ DATA AND FIT A SIMPLE MODEL ###

X, y = fetch_california_housing(return_X_y=True, as_frame=True)

m = Ridge()
m.fit(X,y)

 

Out[3]:

Ridge()

 

In [4]:

perf = PermutationImportance(model=m , output_type="regression")

 

In [5]:

imp_table = perf.make_imp_table(X,y,column_names=list(X))

 

 

100%|██████████| 8/8 [00:00<00:00, 28.34it/s]

 

In [6]:

imp_table

 

Out[6]:

	mean	std
MedInc	1.379051	0.014738
HouseAge	0.028542	0.001399
AveRooms	0.139063	0.002949
AveBedrms	0.186004	0.002765
Population	0.000045	0.000056
AveOccup	0.002938	0.000298
Latitude	1.620343	0.012850
Longitude	1.522178	0.013827

 

In [7]:

perf.plot_permutation_importance(imp_table)

 

 

 

In [8]:

sample_idx = np.random.choice(np.arange(1,len(X)),1000)

 

In [9]:

sample_X , sample_y = X.iloc[sample_idx , : ] , y[sample_idx]

 

In [10]:

sample_imp_table = perf.make_sample_imp_table(sample_X,sample_y,column_names=list(X))

 

 

100%|██████████| 8/8 [00:00<00:00, 40.32it/s]

 

In [11]:

perf.plot_sample_permutatiom_importance(sample_imp_table)

 

 

 

In [12]:

perf.plot_one_sample_importacne_heatmap(sample_imp_table , 0)

 

 

 

In [13]:

perf.plot_sample_importacne_heatmap_betwwen_columns(sample_X , sample_imp_table , 'MedInc' , "HouseAge" , "AveBedrms")

 

 

 

In [14]:

perf.plot_sample_importance_heatmap(sample_X ,sample_imp_table)

 

 

 

Classiftion¶

 

In [15]:

X, y = fetch_openml(name='wine-quality-red', return_X_y=True, as_frame=True)

m = LogisticRegression(C=10, max_iter=10_000)
m.fit(X,y)

 

Out[15]:

LogisticRegression(C=10, max_iter=10000)

 

In [16]:

perf = PermutationImportance(model=m , output_type="classification")

 

In [17]:

imp_table = perf.make_imp_table(X,y,column_names=list(X))

 

 

100%|██████████| 11/11 [00:00<00:00, 29.70it/s]

 

In [18]:

perf.plot_permutation_importance(imp_table)

 

 

 

In [19]:

sample_idx = np.random.choice(np.arange(1,len(X)),1000)
sample_X , sample_y = X.iloc[sample_idx , : ] , y[sample_idx]

 

In [20]:

sample_imp_table = perf.make_sample_imp_table(sample_X,sample_y,column_names=list(X))

 

 

100%|██████████| 11/11 [00:02<00:00,  4.26it/s]

 

In [21]:

perf.plot_sample_permutatiom_importance(sample_imp_table)

 

 

 

In [22]:

perf.plot_one_sample_importacne_heatmap(sample_imp_table , 1)

 

 

 

In [23]:

perf.plot_sample_importance_heatmap(sample_X ,sample_imp_table)

 

 

 

In [24]:

perf.plot_sample_importacne_heatmap_betwwen_columns(sample_X , sample_imp_table , 'fixed_acidity' , "volatile_acidity" , "chlorides")

 

 

 

In [25]:

perf.plot_sample_importance_target_heatmap(sample_X , sample_imp_table,"fixed_acidity",n_bins=20)

 

 

 

In [ ]: