def Titanic():
X = pd.read_csv('datas/train.csv')
# указываем зависимую перменную
y = X['Survived']
#X.head()
# смотрим, как распределены выжившие в зависимости от пола
X[["Sex", "Survived"]].groupby(['Sex'], as_index=False).mean().sort_values(by='Survived', ascending=False)
# удаляем из входов зависимую перменную и незначимые признаки
X.drop(['Survived', 'Name', 'PassengerId', 'Ticket'], axis=1, inplace=True)
X.head()
X.info()
# в поле Cabin много пропусков, удалим и его
X.drop(['Cabin'], axis=1, inplace=True)
X['Embarked'].describe()
# дозаполняем пропуски
X['Age'].fillna(X['Age'].median(), inplace=True)
X['Embarked'].fillna('S', inplace=True)
X.info()
# кодируем поле Embarked методом дамми-кодирования
X = pd.concat([X, pd.get_dummies(X['Embarked'], prefix="Embarked")], axis=1)
# удаляем старое поле Embarked
X.drop(['Embarked'], axis=1, inplace=True)
X['Sex'] = pd.factorize(X['Sex'])[0]
X.info()
# делим выборку на обучающую и тестовую
X_train = X[:-200]
X_test = X[-200:]
y_train = y[:-200]
y_test = y[-200:]
clf = tree.DecisionTreeClassifier(max_depth=5, random_state=21)
clf.fit(X_train, y_train)
clf.score(X_train, y_train)
clf.score(X_test, y_test)
#rfc = RandomForestClassifier(n_estimators=10, max_depth=5, random_state=21)
#rfc.fit(X_train, y_train)
#rfc.score(X_test, y_test)
eli5.explain_weights_sklearn(clf, feature_names=X_train.columns.values)
plot_tree(clf, filled=True)
plt.show()
export_graphviz(clf, out_file='datas/pic.dot')
def test_explain_linear_multilabel(clf):
X, Y = make_multilabel_classification(random_state=42)
clf.fit(X, Y)
res = explain_weights_sklearn(clf)
expl_text, expl_html = format_as_all(res, clf)
for expl in [expl_text, expl_html]:
assert 'y=4' in expl
assert 'x0' in expl
assert 'BIAS' in expl
def test_format_html_options(force_weights, horizontal_layout):
# test options that are not tested elsewhere
X, y = make_regression(n_samples=100, n_targets=3, n_features=10,
random_state=42)
reg = LinearRegression()
reg.fit(X, y)
res = explain_weights_sklearn(reg)
kwargs = dict(
force_weights=force_weights, horizontal_layout=horizontal_layout)
postfix = '_' + '_'.join(
'{}-{}'.format(k, v) for k, v in sorted(kwargs.items()))
print(kwargs, postfix)
# just check that it does not crash
expl = format_as_html(res, **kwargs)
write_html(reg, expl, format_as_text(res), postfix=postfix)
pred_res = explain_prediction_sklearn(reg, X[0])
pred_expl = format_as_html(pred_res, **kwargs)
write_html(reg, pred_expl, format_as_text(pred_res),
postfix='_expl' + postfix)