Your First AutoML Model in Python

The first AutoML run should be small, clear, and easy to inspect. The goal is not maximum score. The goal is to see the whole workflow end to end.

Start with a clean dataset

For a first experiment, choose a tabular dataset with:

one clearly defined target column
a manageable number of rows
a mix of numeric or categorical features you can explain

Keep the first run simple enough that you can inspect every output.

Launch the training run

With mljar-supervised, a minimal workflow looks like this:

from supervised import AutoML

automl = AutoML(mode="Explain")
automl.fit(X_train, y_train)

That one call triggers the training workflow, but the useful part comes after training: inspecting what was produced.

AutoML training progress in Python workflow

Read the generated outputs

After the run finishes, inspect:

leaderboard results
model-by-model metrics
feature importance
saved report artifacts

These outputs help answer practical questions:

Which model performed best?
How large was the gap between top models?
Which features influenced the prediction most?
Is the result strong enough to justify another iteration?

What to do after the first run

Your first AutoML model is a baseline, not the final answer. The next iteration usually focuses on one of four things:

Improve the data

Better target definition, better cleaning, and better feature construction often matter more than another round of blind hyperparameter search.

Adjust the validation

If the validation scheme does not match reality, the score can look better than the actual future performance.

Compare explanation depth

Different AutoML modes trade off speed and explainability. Choose the mode that fits the current stage of the project.

Tighten the objective

Sometimes the real improvement is not a better algorithm. It is a better metric, a better threshold, or a better understanding of business cost.

This is the point where AutoML becomes practical: not when it gives you a number, but when it gives you a structured starting point for the next decision.

Your First AutoML Model