1pip install auto-sklearn

1import autosklearn.classification
2from sklearn.model_selection import train_test_split
3from sklearn.datasets import load_breast_cancer
4import sklearn.metrics
5
6# Load data
7X, y = load_breast_cancer(return_X_y=True)
8X_train, X_test, y_train, y_test = train_test_split(
9    X, y, test_size=0.2, random_state=42
10)
11
12# AutoML classifier
13automl = autosklearn.classification.AutoSklearnClassifier(
14    time_left_for_this_task=300,  # 5 minutes
15    per_run_time_limit=60,         # 1 minute per model
16    n_jobs=-1,
17    ensemble_size=10,
18    memory_limit=4096
19)
20
21# Fit
22automl.fit(X_train, y_train)
23
24# Evaluate
25y_pred = automl.predict(X_test)
26accuracy = sklearn.metrics.accuracy_score(y_test, y_pred)
27print(f"Accuracy: {accuracy:.4f}")
28
29# Show models
30print(automl.leaderboard())

1import h2o
2from h2o.automl import H2OAutoML
3
4# Start H2O
5h2o.init()
6
7# Load data
8df = h2o.import_file("data/train.csv")
9
10# Define features and target
11y = "target"
12x = df.columns
13x.remove(y)
14
15# Split
16train, test = df.split_frame(ratios=[0.8], seed=42)
17
18# Run AutoML
19aml = H2OAutoML(
20    max_runtime_secs=300,
21    max_models=20,
22    seed=42,
23    sort_metric="AUC"
24)
25
26aml.train(x=x, y=y, training_frame=train)
27
28# Leaderboard
29lb = aml.leaderboard
30print(lb)
31
32# Best model
33best = aml.leader
34perf = best.model_performance(test)
35print(perf)
36
37# Predictions
38preds = best.predict(test)

1# SHAP values
2import h2o.explain as h2o_explain
3
4# Explain single prediction
5h2o_explain.explain_row(aml.leader, train, row_index=0)
6
7# Feature importance
8h2o_explain.varimp_heatmap(aml.leader)

1pip install autogluon

1from autogluon.tabular import TabularDataset, TabularPredictor
2
3# Load data
4train_data = TabularDataset('train.csv')
5test_data = TabularDataset('test.csv')
6
7# Train AutoML
8predictor = TabularPredictor(
9    label='target',
10    eval_metric='accuracy',
11    problem_type='binary'
12).fit(
13    train_data,
14    time_limit=300,
15    presets='best_quality'  # or 'medium_quality_faster_train'
16)
17
18# Leaderboard
19predictor.leaderboard(test_data)
20
21# Predict
22predictions = predictor.predict(test_data)
23
24# Feature importance
25predictor.feature_importance(test_data)

1# Best quality (slow, accurate)
2predictor.fit(train_data, presets='best_quality')
3
4# Medium quality (balanced)
5predictor.fit(train_data, presets='medium_quality_faster_train')
6
7# Fast (quick prototyping)
8predictor.fit(train_data, presets='optimize_for_deployment')

1from pycaret.classification import *
2
3# Initialize
4clf = setup(
5    data=train_df,
6    target='target',
7    session_id=42,
8    normalize=True,
9    feature_selection=True
10)
11
12# Compare all models
13best = compare_models()
14
15# Tune best model
16tuned = tune_model(best)
17
18# Ensemble
19ensembled = ensemble_model(tuned, method='Bagging')
20
21# Blend multiple models
22blend = blend_models(top3=[best, tuned, ensembled])
23
24# Finalize
25final = finalize_model(blend)
26
27# Predict
28predictions = predict_model(final, data=test_df)
29
30# Save model
31save_model(final, 'my_model')

1from pycaret.classification import *
2
3# Custom scorer
4from sklearn.metrics import f1_score, make_scorer
5
6custom_scorer = make_scorer(f1_score, average='weighted')
7
8# Setup with custom metric
9clf = setup(
10    data=train_df,
11    target='target',
12    custom_metric=custom_scorer
13)

1from tpot import TPOTClassifier
2from sklearn.model_selection import train_test_split
3
4# Split data
5X_train, X_test, y_train, y_test = train_test_split(
6    X, y, test_size=0.2, random_state=42
7)
8
9# TPOT AutoML
10tpot = TPOTClassifier(
11    generations=10,
12    population_size=50,
13    offspring_size=50,
14    scoring='accuracy',
15    cv=5,
16    random_state=42,
17    n_jobs=-1,
18    verbosity=2
19)
20
21tpot.fit(X_train, y_train)
22
23# Score
24print(f"Score: {tpot.score(X_test, y_test):.4f}")
25
26# Export best pipeline
27tpot.export('best_pipeline.py')

1import pandas as pd
2from autogluon.tabular import TabularDataset, TabularPredictor
3from sklearn.model_selection import train_test_split
4import matplotlib.pyplot as plt
5
6# Load data
7df = pd.read_csv('data.csv')
8
9# Basic EDA
10print(df.info())
11print(df.describe())
12
13# Check target distribution
14print(df['target'].value_counts())
15
16# Split
17train_df, test_df = train_test_split(df, test_size=0.2, random_state=42, stratify=df['target'])
18
19# Convert to AutoGluon format
20train_data = TabularDataset(train_df)
21test_data = TabularDataset(test_df)
22
23# Train with AutoGluon
24predictor = TabularPredictor(
25    label='target',
26    eval_metric='roc_auc',
27    problem_type='binary'
28).fit(
29    train_data,
30    time_limit=600,  # 10 minutes
31    presets='best_quality',
32    verbosity=2
33)
34
35# Results
36print("\n" + "="*50)
37print("RESULTS")
38print("="*50)
39
40# Leaderboard
41lb = predictor.leaderboard(test_data, silent=True)
42print("\nLeaderboard:")
43print(lb)
44
45# Best model performance
46perf = predictor.evaluate(test_data)
47print(f"\nTest Performance: {perf}")
48
49# Feature importance
50importance = predictor.feature_importance(test_data)
51print("\nFeature Importance:")
52print(importance)
53
54# Plot feature importance
55importance.head(20).plot(kind='barh', figsize=(10, 8))
56plt.title('Top 20 Feature Importance')
57plt.tight_layout()
58plt.savefig('feature_importance.png')
59
60# Save predictor
61predictor.save('my_automl_model')
62
63# Load and predict on new data
64predictor = TabularPredictor.load('my_automl_model')
65new_predictions = predictor.predict(new_data)