Features importance

The methode features_importance displays a bar chart representing the sum of absolute contribution values of each feature.

This method also makes it possible to represent this sum calculated on a subset and to compare it with the total population

This short tutorial presents the different parameters you can use.

Content : - Classification case: Specify the target modality to display. - selection parameter to display a subset - max_features parameter limits the number of features

We used Kaggle’s Titanic dataset

from sklearn.ensemble import ExtraTreesClassifier
from sklearn.model_selection import train_test_split

Building Supervized Model

Load Titanic data

from shapash.data.data_loader import data_loading
titanic_df, titanic_dict = data_loading('titanic')
del titanic_df['Name']
y_df=titanic_df['Survived'].to_frame()
X_df=titanic_df[titanic_df.columns.difference(['Survived'])]

titanic_df.head()

	Survived	Pclass	Sex	Age	SibSp	Parch	Fare	Embarked	Title
PassengerId
1	0	Third class	male	22.0	1	0	7.25	Southampton	Mr
2	1	First class	female	38.0	1	0	71.28	Cherbourg	Mrs
3	1	Third class	female	26.0	0	0	7.92	Southampton	Miss
4	1	First class	female	35.0	1	0	53.10	Southampton	Mrs
5	0	Third class	male	35.0	0	0	8.05	Southampton	Mr

Load Titanic data

from category_encoders import OrdinalEncoder

categorical_features = [col for col in X_df.columns if X_df[col].dtype == 'object']

encoder = OrdinalEncoder(
    cols=categorical_features,
    handle_unknown='ignore',
    return_df=True).fit(X_df)

X_df=encoder.transform(X_df)

Train / Test Split + model fitting

Xtrain, Xtest, ytrain, ytest = train_test_split(X_df, y_df, train_size=0.75, random_state=7)

clf = ExtraTreesClassifier(n_estimators=200, random_state=79).fit(Xtrain,ytrain['Survived'])

First step: You need to Declare and Compile SmartExplainer

from shapash import SmartExplainer

response_dict = {0: 'Death', 1:' Survival'}

xpl = SmartExplainer(
    model=clf,
    preprocessing=encoder,      # Optional: compile step can use inverse_transform method
    features_dict=titanic_dict, # Optional parameters
    label_dict=response_dict    # Optional parameters, dicts specify labels
)

xpl.compile(x=Xtest, y_target=ytest)

INFO: Shap explainer type - <shap.explainers._tree.TreeExplainer object at 0x7f3f2fd5b730>

Display Feature Importance

xpl.plot.features_importance()

Multiclass: Select the target modality

Features importances sum and display the absolute contribution for one target modality. you can change this modality, selecting with label parameter:

xpl.plot.features_importance(label=‘Death’)

with label parameter you can specify target value, label or number

Focus and compare a subset

selection parameter specify the subset:

sel = [581, 610, 524, 636, 298, 420, 568, 817, 363, 557,
       486, 252, 390, 505, 16, 290, 611, 148, 438, 23, 810,
       875, 206, 836, 143, 843, 436, 701, 681, 67, 10]

xpl.plot.features_importance(selection=sel)

Tune the number of features to display

Use max_features parameter (default value: 20)

xpl.plot.features_importance(max_features=3)

Understand local effect

This plot allows us to observe how the importance of features varies across different subpopulations. For instance, we can see that in certain subpopulations, the Port of Embarkation has a greater impact than the Ticket Class, highlighting the local variations in feature significance.

xpl.plot.features_importance(mode='global-local', max_features=10, zoom=True)

In the plot below, we observe the same effect as before. For example, in certain subpopulations, the Port of Embarkation has a greater impact than the Ticket Class. This offers another way to visualize feature importance both locally and globally.

When the curves cross each other, it indicates that one feature has a higher local effect in a specific subpopulation, but a lower global impact across the entire dataset. On the other hand, if a curve consistently remains higher than another, it signifies that the feature is more important both globally and locally.

After this initial analysis, you can use the contribution plot to gain deeper insights into how a particular feature influences the model’s predictions.

xpl.plot.features_importance(mode='cumulative', normalize_by_nb_samples=True, degree=-0.7, zoom=True)