Please, explain! Interpretable machine learning based on ranking data

Name: Please, explain! Interpretable machine learning based on ranking data
Start: 2019-12-10T18:00:00+01:00
End: 2019-12-10T21:30:00+01:00
Location: Navan Amsterdam

Georganiseerd door Janine K. en Sarah S.

Ontmoet de groep

R-Ladies Amsterdam

Nog geen beoordelingen

Details

Opening the black box is a fertile area of research today. It is apparent to anyone who has ever worked on a machine learning project that in order to realise a wide spread acceptance delivering a high accuracy model is not enough. Stakeholders are demanding transparency in how and why an algorithm is making specific predictions. A variety of approaches can be used to shed light on how algorithms make their decision: from creating alternative and more transparent algorithms that can compete with the black-box one to using robust statistical techniques. During this talk we will explore different methods for unboxing the black-box model, their pros and cons and how I have used them in TripActions.
We will cover the following topics:

Importance of human interpretable machine learning and criteria for model interpretation methods.
Challenges and limitations of traditional techniques for model interpretation
Hands-on example on the latest model interpretation frameworks on a supervised learning example. Using the iml package we will cover permutation importance, SHAP and LIME.

This workshop is brought to you by Teodora Vrabcheva. :)

PROGRAM

18.00-18.30 Welcoming :)
18.30- 18.50 Introduction to R-Ladies & Tripactions
18.50- 19.30- Workshop part 1
19.30- 19.45 Break
19.45- 20.30 Workshop part 2
20.30- 21.30 Socialising

ABOUT TEODORA

https://www.linkedin.com/in/teodora-vrabcheva-8b841826/

My name is Teodora and I am a data scientist with background in statistics. I have 8 years of experience in data analytics and data science. I started my career in London working as a statistician in healthcare and marketing research. For the past 3 years I am based in Amsterdam and currently I work for TripActions where my main focus is recommender systems.

REQUIREMENTS

Basic knowledge of machine learning/random forest is sufficient. Installation of `iml`- package.

ABOUT TripActions

TripActions is an end-to-end business travel platform making business travel as easy and delightful as possible for road warriors. With unrivaled flight, lodging and rental car inventory and global 24/7 365 proactive live human support we maintain 97% traveler satisfaction. In the data team we are responsible for building a scalable data warehouse, reporting and advanced analytics, and providing personalised customer experience through machine learning.

CODE OF CONDUCT

We expect all attendees to abide by our Code of Conduct. Anyone who violates this community's Code of Conduct will be refused entry, expelled, and/or any other action deemed appropriate by the Organizer, regardless of event registration, meetup membership or other condition. We are emphatically queer and trans-friendly. This group was created so those who identify as women have a comfortable place to learn.