Data is today’s clay and the newest killer applications are data products.

In this workshop, you will learn about the anatomy of a data product on Spark and H2O. Then we’ll show you how to mold your data in order to build a pipeline and a real-time ML system by building a complete loan interest rate prediction product. Our clay is going to be a public Lending Club dataset, and we will create a real-time Spark Streaming application that is both predictive and deploys library models.

The goal is to produce borrower interest rates that are comparable or better than human-led predictions.

Key Take Aways:

• Clean and transform datasets in Sparkling Water
• Join varying datasets (text and time series) by defining conformed dimensions
• Use MLlib to implement word2vec for NLP and H2O for Gradient Boosting Machine in order to produce a scoring engine
• Integrate the scoring engine from Sparkling Water models into Spark Streaming
• Produce real-time scoring and predictions
• Create a pipeline of ensemble models – retire and promote them based on risk and domain
• Deploy smarter applications on Spark and Cloud