AutoGluon 1.2 tutorial: classification and regression

This Python tutorial uses AutoGluon 1.2 to train models for binary classification and regression. The presenter describes it as an open source machine learning toolkit developed by Amazon. The walkthrough uses a notebook in VS Code, though the presenter says viewers can use another editor, and covers installation with pip on Mac and Windows.

The classification example uses a prepared dataset with loan default as the target and inputs such as home ownership, credit score and income. The presenter samples 80% of the data for training and keeps the remaining rows for testing. A TabularPredictor fit call starts training, after which the tutorial inspects the saved model files and compares test and validation accuracy in a leaderboard. The demonstrated run produces 30 models; that count describes this example rather than every training run.

The presenter then generates predictions without the target column and evaluates them against the known labels. Although the introduction describes automated hyperparameter tuning, the presenter characterizes this run as using base parameters and recommends further tuning to improve performance.

The regression example predicts house prices using bedrooms, bathrooms, square footage and garage information. It repeats the data split, explicitly sets the problem type to regression, and inspects the resulting leaderboard and saved models.