Python) Catboost 모델링 및 Shap으로 분석해보기

 

In [1]:

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

 

/tmp/ipykernel_2227596/3510566465.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

 

 

In [2]:

import shap

# train XGBoost model
X,y = shap.datasets.adult()

In [5]:

import pandas as pd
import sys 
ROOT_DIR = "./../"
sys.path.append(ROOT_DIR)
import shap
import matplotlib.pyplot as plt ## 시각화
import matplotlib.font_manager as fm## 시각화
import matplotlib
import catboost as cb
import numpy as np
import warnings ## 에러 방지
warnings.filterwarnings(action='ignore') 
## 한글 폰트 (나눔) 적용하는 코드
plt.rc('axes',unicode_minus=False)
# plt.rcParams["font.family"] = 'NanumGothic' 
# matplotlib.rcParams['axes.unicode_minus'] = False
# plt.rc('font', family='NanumGothic') 
# path = '/usr/share/fonts/truetype/nanum/NanumGothicCoding.ttf'
plt.rc('font', family='NanumBarunGothic') # 폰트 설정
# plt.rc("font", family = "Malgun Gothic")
plt.rc('axes', unicode_minus=False) # 마이너스 폰트 설정
from IPython.display import set_matplotlib_formats
set_matplotlib_formats("retina")

In [18]:

X,y = shap.datasets.adult(display=True)
cat_features = list(X.select_dtypes("category"))
train_dataset = cb.Pool(X, y,cat_features=cat_features)
model = cb.CatBoostClassifier(verbose=False)
grid = {'iterations': [100, 150, 200],
        'learning_rate': [0.03, 0.05,0.07,0.1],
        'depth': [6, 8,10],
        'l2_leaf_reg': [0.2, 0.5, 1, 3]}
# model.grid_search(grid, train_dataset)
model_result = model.randomized_search(grid, train_dataset,n_iter=2,
                        plot=False,
                        verbose=False,partition_random_seed=0)

 

bestTest = 0.2842787463
bestIteration = 99


bestTest = 0.2856006001
bestIteration = 149

Training on fold [0/3]

bestTest = 0.2923580878
bestIteration = 98

Training on fold [1/3]

bestTest = 0.2938665719
bestIteration = 99

Training on fold [2/3]

bestTest = 0.2807298121
bestIteration = 97

In [19]:

pd.DataFrame([model.get_params()])

Out[19]:

	verbose	depth	l2_leaf_reg	iterations	learning_rate
0	False	8	1	100	0.07

In [20]:

sorted_feature_importance = model.feature_importances_.argsort()
plt.barh(np.array(list(X))[sorted_feature_importance], 
        model.feature_importances_[sorted_feature_importance], 
        color='turquoise')
plt.xlabel("CatBoost Feature Importance")

Out[20]:

Text(0.5, 0, 'CatBoost Feature Importance')

 

In [21]:

approximate = model.get_feature_importance(train_dataset, type="ShapValues", shap_calc_type="Approximate")[:, :-1]
shap.summary_plot(approximate, X,plot_type="bar")

 

In [22]:

approximate = model.get_feature_importance(train_dataset, type="ShapValues", shap_calc_type="Exact")[:, :-1]
shap.summary_plot(approximate, X,plot_type="bar")

 

In [23]:

approximate = model.get_feature_importance(train_dataset, type="ShapValues", shap_calc_type="Regular")[:, :-1]
shap.summary_plot(approximate, X,plot_type="bar")

 

In [24]:

shap.dependence_plot("Relationship", approximate, X)

 

In [25]:

explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(cb.Pool(X, y, cat_features=cat_features))

In [26]:

shap.initjs()
sample_idxs = [0,1,2]
shap.force_plot(explainer.expected_value, shap_values[sample_idxs,], X.iloc[sample_idxs,:])

 

Out[26]:

Visualization omitted, Javascript library not loaded!
Have you run `initjs()` in this notebook? If this notebook was from another user you must also trust this notebook (File -> Trust notebook). If you are viewing this notebook on github the Javascript has been stripped for security. If you are using JupyterLab this error is because a JupyterLab extension has not yet been written.

In [27]:

shap.summary_plot(shap_values, X, plot_type="bar")

 

In [28]:

shap.summary_plot(shap_values, X)

 

In [29]:

cols = list(X.select_dtypes("category"))
import category_encoders as ce 
encoder = ce.OrdinalEncoder(cols=cols)

In [30]:

shap_values_v2 = explainer(
    cb.Pool(X, y, cat_features=cat_features))
exp =shap.Explanation(
    values=shap_values_v2.values ,
    base_values= shap_values_v2.base_values,
    data=encoder.fit_transform(X).values,
    feature_names=list(X),
    display_data=X.values
    )

In [31]:

for col in cols : 
    shap.plots.scatter(exp[:,col])

 

 

 

 

 

 

 

In [32]:

shap_values_v2 = explainer(cb.Pool(X, y, cat_features=cat_features))
exp =shap.Explanation(
    values=shap_values_v2.values , 
    base_values= shap_values_v2.base_values,
    data=X.values,
    feature_names=list(X),
    display_data=X.values
    )

In [34]:

import matplotlib.pyplot as plt
from shap.plots import waterfall
waterfall(exp[2],show=True,max_display=22)

 

In [35]:

shap.plots.heatmap(exp[0:500],max_display=22)

 

In [ ]: