autogluon_feature_generator_quickstart_tabular_transform.py

1import pandas as pd
2from autogluon.features.generators import AutoMLPipelineFeatureGenerator
3
4# 1. Create a sample raw dataset
5data = pd.DataFrame({
6    'age': [25, 30, 45, 22],
7    'city': ['San Francisco', 'New York', 'New York', 'London'],
8    'salary': [50000, 80000, 120000, 45000],
9    'datetime': ['2020-01-01', '2020-02-15', '2021-03-10', '2022-05-20']
10})
11
12# 2. Initialize the FeatureGenerator
13# AutoMLPipelineFeatureGenerator is the default used in TabularPredictor
14generator = AutoMLPipelineFeatureGenerator()
15
16# 3. Fit the generator and transform the data
17# This automatically handles missing values, date parsing, and categorical encoding
18transformed_data = generator.fit_transform(X=data)
19
20# 4. View results
21print("Original Columns:", data.columns.tolist())
22print("Transformed Columns:", transformed_data.columns.tolist())
23print("\nTransformed Data Snapshot:")
24print(transformed_data.head())