Recipes in H2O Driverless AI
Overview
Driverless AI + Your Recipes = A Truly Extensible AI Platform
Recipes are customizations and extensions to the Driverless AI platform. These can be custom machine learning models, transformers, or scorers (classification or regression), written in Python. Data scientists can bring their own recipes or leverage the open-source recipes available by the community and curated by H2O.ai data science experts.
Benefits of Recipes
- Flexibility, extensibility and customizations built into the Driverless AI platform
- New open source recipes built by the data science community, curated by Kaggle Grand Masters @ H2O.ai
- Data scientists can focus on domain-specific functions to build customizations.
- 1-click upload of your recipes – models, scorers and transformations
- Driverless AI treats custom recipes as first-class citizens in the automatic machine learning workflow.
- Every business can have a recipe cookbook for collaborative data science within their organization
How Driverless AI Recipes Work
What data scientists can do with Driverless AI Recipes
Models
An insurance provider could recommend the right insurance product to its customers by building a gradient boost model based on the CatBoost algorithm recipe.
Transformers
A retailer could forecast annual sales based on seasonality, weather and Ramadan holidays for its stores in Saudi Arabia.
Scorers
An oil and gas company could predict oil-well output by developing a time-series regression model and use the Mean Absolute Scaled Error recipe to measure the accuracy of the forecasted output.
A media television company could use the historic mean time-series recipe to smooth out the distribution graph of the # of viewers throughout a busy sports tournament by breaking the entire season in discrete time windows.
A network analytics company can improve its anti-IP-spoofing functionality by parsing IP addresses and checking if any of the properties looks like a spam or a DoD attack.
A transportation and logistics company could use the precision-recall classification recipe to accurately predict the on-time delivery of perishable products by tweaking the threshold depending upon weather conditions.
A sports and entertainment event broadcast company could do NLP for specific Twitter handles and hashtags, and classify sentiments of every event using the Intel DAAL models, enabling them to be smart about targeting demographics.
A financial services company could calculate the historical volatility of markets in a configurable rolling time-window on time-series financial data.
A marketing and lead-gen company could improve the effectiveness of its products by accurately predicting its ROI for its customers, using the marketing campaign scorer recipe.
Build Your Own Custom Recipe Tutorial
This tutorial is for H2O Driverless AI; you will learn how to build and troubleshoot your own custom transformer, scorer and model recipes.
In this tutorial, we will take a deeper dive into the process of custom recipe building to enhance Driverless AI. We will build three recipes: a transformer, a model and a scorer using Visual Studio Code text editor.
