useR Tallinn @ Coop Bank
Details
Happy New Year to everyone, and welcome to our first meeting of the year, at Coop Bank. Our host, Mart Kaska, has been using R to support anti-money laundering (AML) efforts at the bank. He will give us an overview of the role R plays in his work, and in his talk will discuss how he's worked to support conventional AML analysis with more quantitative insights. In the second talk I'm going to speak about how, and why, to write R applications with a RESTful web API using plumber and Docker.
Program
17:30 - Arrival and snacks
18:00 - Introductory remarks by Mart Kaska, Coop Bank
18:15 - Start of presentations (~20m each)
First talk - Mart Kaska, Data Scientist at Coop Bank, "How to catch bad guys with R?"
Conventional AML analysis relies on analysts' domain knowledge on what kinds of customer transactions are likely to indicate illegal behavior. Mart has been using R to prototype basic quantiative models of fraudulent activity, both in order to test whether it can help analysts prioritize high risk cases, and also to check whether model-based approaches can identify problematic cases conventional analysis misses. As is often the case when expert knowledge meets math, some challenges arise in bridging two very different approaches to AML.
Second talk - Andreas Beger, Data Scientist at Predictive Heuristics, "Writing R web APIs with plumber and Docker"
Two problems arise when moving from analysis that run on your own laptop to "live" functionality in a diverse team: how to allow non-R users to interface with R-based functionality, and second, how to make your code run on someone else's computer (or server). The plumber package provides a solution for the first problem by turning R code and functions into a web application with a RESTful API. Docker is a solution to the second problem by packaging all of this into a container that can be setup anywhere where Docker is installed, regardless of the underlying hardware.
Third talk – Simon Wenkel, Machine Learning Engineer, “R – Python – Julia. Insights into old and new languages for data science and machine learning and implications for their use in (high performance) production environments”
Using interpreted languages has many advantages for prototyping/exploratory work but certain disadvantages for production usage. To overcome some of these issues, we’ll
- have a look at these languages and how there are implemented,
- have a look under the hood of common data science/machine learning stacks in these 3 languages,
- see how to speed them up without writing extensions in C/C++/Fortran,
- have a look at things to consider when deploying them in production environments.
Logistics
The meeting will take place at Coop Bank, Narva mnt 4. This is right next to the Hobujaama tram stops. In case you are driving, there are a small number of free parking spots next to the building. Once inside, tell the front desk person you came to the meetup, go past the turnstiles on the right and up to the 4th floor conference room.
Snacks and refreshments will be available, and you can bring food if you'd like.
Meeting GitHub repo
useR Tallinn @ Coop Bank