Automated Data Prep in H2O Driverless AI
- Predictive AI
- H2O Driverless AI
- Enterprise h2oGPTe
This lesson walks through how loading data into Driverless AI immediately triggers automated profiling, covering missing value detection and distribution analysis so data quality issues surface before modeling begins. It shows how the platform wrangles data through categorical encoding, imputation, normalization, and feature transformations, all embedded directly into the scoring pipeline so training and inference use identical logic. The lesson also covers automatic lag feature generation for time series data, and notes that synthetic structured data can be created through prompts using Enterprise h2oGPTe agents.
Key takeaways
- Why automated data profiling on load surfaces missing values and distribution issues before modeling starts
- How embedding preprocessing logic directly into the scoring pipeline avoids mismatches between training and inference
- What automatic lag features and test set augmentation add specifically for time series datasets
- How Enterprise h2oGPTe agents can generate new synthetic structured data from natural language prompts
Start owning your AI