Choosing the Right AI Model Using Business Value
- Predictive AI
- H2O Driverless AI
This lesson walks through choosing a machine learning model based on business value rather than pure accuracy, using a customer churn project as the example. It covers Driverless AI's Business Value Calculator, which estimates profit for each model at its optimized threshold based on the cost of being right or wrong about a customer churning. It also shows how expert settings let you tune interpretability, accuracy, and training time, contrasting a churn model that only needs accuracy with a regulated use case that also needs to be explainable. It closes with tips for edge deployment, including MOJO size and latency scoring.
Key takeaways
- How the Business Value Calculator estimates profit for each model at its own optimized decision threshold
- Why a regulated use case needs higher interpretability even if it costs some accuracy compared to a churn model
- How expert settings let you trade off training time, accuracy, and interpretability for a specific business case
- What to check before deploying to an edge device: MOJO size and latency scoring across candidate models
Start owning your AI