Data Science Liege MEETUP #2

Abstract

Combining Jupyter, Docker and Dask to build a scalable machine learning environment –  Dr. Loic Quertenmont (n-side)

In this presentation, we will discuss how we have combined three traditional tools in order to get a production-ready and scalable machine learning environment. The advantages of such a setup will be discussed as well as it’s daily usage at N-Side for our electricity price forecasting platform.

From Robotic Process Automation to Artificial Intelligence –  Tolga Bayrak (Accenture)

“Artificial intelligence”, “Automation” and “Robotics” are buzzwords that are thrown around a lot lately. But what are they really and how do they relate to each other? Are they really being applied in the enterprise? If so, how does one go about them?

Through this presentation, I will start by introducing the basic concepts of Robotic Process Automation (or “RPA”), which has seen a spectacular success in the enterprise during the last few years – just look at Blue Prism’s stock value evolution over the same period (hint: it’s over +1000%).

I will then circle back to artificial intelligence, and show how both technologies relate to each other in what we call “intelligent automation”. Finally, I’ll describe the intelligent automation journey, or how any organization can implement and benefit from these game-changing technologies.

Consumer Behavior Analytics for Household Goods: A Procter&Gamble case – Dr. Leïla Rebbouh (LR Physics)

P&G and Electrolux have recently decided to partner in the implementation of an e-commerce and analytic tools to study customer behavior. At first, we propose to focuse on the effect of market bundling and customer loyalty in particular. Then we will see how an e-commerce can materialize as a programmable button. Finally, we will project in the near future where the washing machine will order its own washing product.

Usage du scraping et des NLP pour récolter, observer et reproduire des conditions générales d’assurance – Aurélien Couloumy (Reacfin)

Insurers produce today more and more qualitative content about regulation, marketing or technical aspects. It becomes critical to equip these parties of new methods to analyze and exploit all this valuable information. One of the first key application than can be valuable for competitor watch is the understanding of insurance general conditions. In this presentation we will shortly introduce scraping techniques we use to collect these document and manage them efficiently. Then we will present characterizing approaches we select to analyze and compare contents using NLP. We will also give some graphical results to show main criteria’s and tendencies through a simple R Shiny interface. Finally we will propose drivers to rebuild a state of art insurance general condition using previous discussions.