H2O.ai Prague Meetup #3 - Booking.com Use Case, ML Pipeline & Explainable AI

Dear Makers,

We are hosting our third meetup on 16th May.

Agenda:
- Doors Open at 6pm. Refreshments + Networking until 6:30pm.
- Welcoming Remarks by H2O.ai team
- Three tech talks by Booking.com & H2O.ai

===

Talk 1:
Water you talking about?

Booking.com has been successfully working with H2O.ai for several years now. We would like to share some great insights that we learned from this collaboration. The topics we will cover are:

Why scalable feature transformation production pipelines are important and how to best build them.
Target encoding and how to efficiently implement it.
Why feature importance for tree models is wrong and how to fix it.

Bios:
Ivana Rebic is a data scientist at Booking.com where she is currently working on machine learning solutions for various business departments.
Before that she worked on product recommendations for hotels as well as building scalable business reports within Booking.com.
She has master degree in pure mathematics and enjoys singing

Santi is a senior developer at booking.com with a MSc in Engineering of Computer Systems from the Politecnico di Milano. He doesn't have much of a background in automated learning yet, but he does have years of only slightly painful experience when it comes to crunching raw business data into reports for humans (and machines!) to process."

===

Talk 2:
Enhancing Spark Pipeline API with H2O Models, Random Grid Search and Automatic Machine Learning using Sparkling Water by Martin Barus

Learn more about how you can integrate large scale data preprocessing with Machine Learning using Sparkling Water. Sparkling Water enables training H2O-3 models leveraging Apache Spark clusters in a distributed manner. It also allows for using trained H2O-3 and Driverless AI models inside Apache Spark. We will demonstrate model training together with hyper-parameter tuning (Cartesian and Random GridSearch with time constraint) of various Algorithms (GBM, GLM), using AutoML - training meta model combining different algorithms, hyper-parameter search and stacking (Ensemble method) all using Spark Pipeline API. We will also demonstrate how trained Driverless AI models can be used for predictions using the same Spark Pipeline API.

Bio:
Martin holds Computer Science Master degree from Czech Technical University in Prague with knowledge engineering specialization. He spent a year as a Master exchange student at the University of Wisconsin-Madison, studying pattern recognition, image processing and machine learning. He is passionate about solving real problems of the physical world using applied mathematics and predictive modeling. Martin is a Customer Data Scientist at H2O.ai

===

Talk 3:
Explainable AI with H2O Driverless AI's machine learning interpretability module by Martin Dvorak

Explainable AI is in the news, and for good reason. Financial services companies have cited the ability to explain AI-based decisions as one of the critical roadblocks to further adoption of AI for their industry. Transparency, accountability, and trustworthiness of data-driven decision support systems based on AI and machine learning are serious regulatory mandates in banking, insurance, healthcare, and other industries. From pertinent regulations, to increasing customer trust, data scientists and business decision makers must show AI-based decisions can be explained. H2O Driverless AI does explainable AI today with its machine learning interpretability (MLI) module. This capability in H2O Driverless
AI employs a unique combination of techniques and methodologies to explain the results of both Driverless AI models and external models.

Bio:
Martin is a passionate software engineer and RESTafarian who is interested in machine learning, VM construction and knowledge management. He holds Master degree in Computer Science from Charles
University Prague with specializations in compilers, operating systems and AI/ML. Martin is a backend engineer on the MLI project at H2O.ai