Dabl

import dabl
import warnings
warnings.filterwarnings('ignore')

Classification

import pandas as pd
train_data = pd.read_csv('./Titanic/train.csv')
X = train_data.drop(["Survived", "PassengerId"], axis=1)
y = train_data.Survived

model = dabl.SimpleClassifier(random_state=0, verbose=1)
model.fit(X, y)

Running DummyClassifier()
accuracy: 0.616 average_precision: 0.384 roc_auc: 0.500 recall_macro: 0.500 f1_macro: 0.381
=== new best DummyClassifier() (using recall_macro):
accuracy: 0.616 average_precision: 0.384 roc_auc: 0.500 recall_macro: 0.500 f1_macro: 0.381

Running GaussianNB()
accuracy: 0.424 average_precision: 0.737 roc_auc: 0.827 recall_macro: 0.529 f1_macro: 0.354
=== new best GaussianNB() (using recall_macro):
accuracy: 0.424 average_precision: 0.737 roc_auc: 0.827 recall_macro: 0.529 f1_macro: 0.354

Running MultinomialNB()
accuracy: 0.763 average_precision: 0.783 roc_auc: 0.831 recall_macro: 0.742 f1_macro: 0.746
=== new best MultinomialNB() (using recall_macro):
accuracy: 0.763 average_precision: 0.783 roc_auc: 0.831 recall_macro: 0.742 f1_macro: 0.746

Running DecisionTreeClassifier(class_weight='balanced', max_depth=1)
accuracy: 0.787 average_precision: 0.632 roc_auc: 0.767 recall_macro: 0.767 f1_macro: 0.771
=== new best DecisionTreeClassifier(class_weight='balanced', max_depth=1) (using recall_macro):
accuracy: 0.787 average_precision: 0.632 roc_auc: 0.767 recall_macro: 0.767 f1_macro: 0.771

Running DecisionTreeClassifier(class_weight='balanced', max_depth=5)
accuracy: 0.788 average_precision: 0.795 roc_auc: 0.838 recall_macro: 0.790 f1_macro: 0.781
=== new best DecisionTreeClassifier(class_weight='balanced', max_depth=5) (using recall_macro):
accuracy: 0.788 average_precision: 0.795 roc_auc: 0.838 recall_macro: 0.790 f1_macro: 0.781

Running DecisionTreeClassifier(class_weight='balanced', min_impurity_decrease=0.01)
accuracy: 0.800 average_precision: 0.788 roc_auc: 0.848 recall_macro: 0.787 f1_macro: 0.788
Running LogisticRegression(C=0.1, class_weight='balanced', max_iter=1000)
accuracy: 0.783 average_precision: 0.819 roc_auc: 0.851 recall_macro: 0.775 f1_macro: 0.772
Running LogisticRegression(class_weight='balanced', max_iter=1000)
accuracy: 0.788 average_precision: 0.816 roc_auc: 0.851 recall_macro: 0.782 f1_macro: 0.778

Best model:
DecisionTreeClassifier(class_weight='balanced', max_depth=5)
Best Scores:
accuracy: 0.788 average_precision: 0.795 roc_auc: 0.838 recall_macro: 0.790 f1_macro: 0.781

SimpleClassifier(random_state=0)

Regression

data = pd.read_csv('housing.csv')
X = data.drop(['median_house_value'], axis = 1) # DataFrame
Y = data.median_house_value

model = dabl.SimpleRegressor(random_state=0, verbose=1)
model.fit(X, Y)

Running DummyRegressor()
r2: -0.000 neg_mean_squared_error: -13316097479.195
=== new best DummyRegressor() (using r2):
r2: -0.000 neg_mean_squared_error: -13316097479.195

Running DecisionTreeRegressor(max_depth=1)
r2: 0.307 neg_mean_squared_error: -9229835649.944
=== new best DecisionTreeRegressor(max_depth=1) (using r2):
r2: 0.307 neg_mean_squared_error: -9229835649.944

Running DecisionTreeRegressor(max_depth=5)
r2: 0.624 neg_mean_squared_error: -5008235435.627
=== new best DecisionTreeRegressor(max_depth=5) (using r2):
r2: 0.624 neg_mean_squared_error: -5008235435.627

Running Ridge(alpha=10)
r2: 0.648 neg_mean_squared_error: -4692624257.386
=== new best Ridge(alpha=10) (using r2):
r2: 0.648 neg_mean_squared_error: -4692624257.386

Running Lasso(alpha=10)
r2: 0.648 neg_mean_squared_error: -4690203117.089
=== new best Lasso(alpha=10) (using r2):
r2: 0.648 neg_mean_squared_error: -4690203117.089


Best model:
Lasso(alpha=10)
Best Scores:
r2: 0.648 neg_mean_squared_error: -4690203117.089

SimpleRegressor(random_state=0)

Tools

titanic_clean = dabl.clean(X, verbose=0)

# detect types
types = dabl.detect_types(X)
types

	continuous	dirty_float	low_card_int	categorical	date	free_string	useless
Pclass	False	False	False	True	False	False	False
Name	False	False	False	False	False	True	False
Sex	False	False	False	True	False	False	False
Age	True	False	False	False	False	False	False
SibSp	False	False	True	False	False	False	False
Parch	False	False	True	False	False	False	False
Ticket	False	False	False	False	False	True	False
Fare	True	False	False	False	False	False	False
Cabin	False	False	False	False	False	True	False
Embarked	False	False	False	True	False	False	False

# plot
dabl.plot(train_data, 'Survived')

Target looks like classification
Linear Discriminant Analysis training set score: 0.589

dabl.explain(model)

Depth: 5
Number of leaves: 27

Reference

• Dabl API