Practical AutoML with Python

What machine learning is, what a model is, what features and targets are, and why ML is useful for domain experts.

Why manual machine learning is hard, what AutoML automates, and what still needs human thinking.

Classification, regression, probabilities, numeric predictions, and real-world examples.

Domain experts, researchers, analysts, doctors, HR specialists, marketers, engineers, students, and business users.

A simple explanation of Python as a tool for data and machine learning.

Different ways to work with Python: scripts, notebooks, and MLJAR Studio.

How to install everything needed to run the examples.

How to organize datasets, notebooks, reports, and trained models.

How tabular data works and how to select input columns and the target column.

Adult Income, Employee Attrition, House Prices, Insurance, Bank Marketing, and Wine.

Why we need to test models on data that was not used during training.

Missing values, categorical columns, duplicated rows, wrong data types, and data leakage.

Train the first MLJAR AutoML model with the Insurance dataset.

Understand the leaderboard, best model, metrics, and output folder.

Use a trained model to make predictions for new data.

Save a model and use it again later.

Regression project with MAE, RMSE, and R².

A more realistic regression project with many input columns.

Binary classification, probabilities, accuracy, and AUC.

HR example with model interpretation and responsible use.

Precision, recall, and choosing the right business metric.

Classification with more than two possible classes.

What files MLJAR AutoML creates and what each file is for.

How to compare models and select the best one.

Accuracy, AUC, F1, MAE, RMSE, R², and when to use each metric.

How to check if the model is learning, overfitting, or not improving.

How to understand which columns are most important.

How to explain individual predictions.

How to understand classification mistakes.

How to understand prediction errors in regression problems.

Best mode for learning, understanding data, and generating explanations.

Best mode for practical use when we want strong models.

Best mode when we want the best possible score and have more time.

Advanced mode for deeper hyperparameter tuning.

A practical guide for choosing the right mode.

Control training time with total_time_limit and model_time_limit.

Select which models AutoML should train.

How combining models can improve prediction quality.

How MLJAR AutoML searches for better model settings.

What feature engineering MLJAR AutoML can do automatically.

Optimize the model for your own business or research metric.

Use train/test split, cross-validation, and custom validation settings.

One of the most important ML mistakes and how to avoid it.

Why models can be unfair and why domain experts should care.

How to check model fairness with sensitive features.

Why machine learning should support people, not blindly replace them.

Make predictions for many rows from a CSV file.

Make one prediction for one person, house, customer, or case.

Create a web app for a trained MLJAR AutoML model.

Let other people use the model without Python.

Move from experiment to something useful for a team, company, or research project.

Wrong target, leakage, bad validation, wrong metric, and blind trust in model results.

A checklist to use before trusting or sharing a model.

APIs, databases, monitoring, retraining, time series, text, and production deployment.

References organized by chapter, with links to papers and official project pages.