Revolutionize Your Data Science Workflows with H2O.ai: The Ultimate Automation Guide!
- Predictive AI
- H2O Driverless AI
Overview
Andrea, an H2O.ai instructor, leads this Driverless AI starter course from the H2O.ai University certification program, walking through the platform end to end using a real credit-score dataset. The session covers setting up a free Aquarium trial account, importing and visualizing data, splitting into train and test sets, and launching a supervised classification experiment using the accuracy, time, and interpretability settings. It also covers reading the results: the ROC curve, the machine learning interpretability report (Shapley values, partial dependence, decision tree surrogate), model diagnostics, and the auto-generated documentation.
Key takeaways
- How to spin up a free, two-hour Aquarium lab account to follow along with Driverless AI without a company license.
- How to import a dataset (a 100,000-row credit-score dataset from H2O's public S3 bucket), review descriptive statistics and column details, and generate automatic visualizations, including an interactive correlation graph.
- How to split data into training and test sets, including stratified sampling on the target column, then launch a supervised classification experiment and tune the accuracy, time, and interpretability knobs.
- How Driverless AI automates feature engineering, model tuning, and selection through an evolutionary process, and how to add custom scoring recipes pulled from the H2O GitHub repository.
- How to read the results: the ROC curve and confusion matrix, then the MLI report's Shapley values, partial dependence plots, and decision tree surrogate model, including a worked example where a misleading "name" feature exposes a data quality issue (repeat applicants) rather than a real predictive signal.
- How to download the auto-generated documentation (AutoDoc) and the Python/MOJO scoring pipelines, and organize experiments into a shareable project.
Start owning your AI