Predictive models at the 2nd meetup anniversary

Hello Sparkers!

We’re glad to announce that this month we turn two years old! Isn’t there a better way than celebrating it with another Meetup? This time we’re inviting Diego García and Tom van der Weide. You’ll probably know them because they’re quite active in this community.

The event will be held on June 7th at Trovit Search offices. We’ll start at 19:00 and after getting inspired by the two use cases we’ll do a networking session (thanks again to Trovit for providing us with the venue, food and beer).

A Segmentation of Water Consumption with Apache Spark, by Diego García

Abstract:
Automatic Meter Reading (AMR) systems are being deployed in many cities to remotely collect water consumptions at higher frequencies (hourly instead of bimonthly) to obtain a better insight of how the water is being consumed by the citizens and industries. This presentation will show a Spark data pipeline to obtain consumption patterns and cluster them by similarity. In this way, a Customer Profiling allows the Water Utilities to achieve different goals, that will be discussed.

Bio:
Diego García, M.S. in Computer Science and M.S. in Automatics and Robotics. I am pursuing Industrial PhD thesis analyzing the contribution of Big Data technologies to the water networks, which is part of a doctorate program initiative of the Generalitat de Catalunya with the aim of improving the synergies between Industry and University. In my case with Aigües de Barcelona and the Specific Research Center (CER) “Monitoring, Safety and Automatic Control” (CS2AC-UPC) of UPC.

Scalable predictive pipelines with Scala and Spark, by Tom van der Weide

Abstract:
Schibsted owns a large number of marketplaces and news websites. For these global sites we are building a platform where advertisers can target ads to certain segments of users, using attributes such as age, gender, and interests. In the user modeling team, we are building predictive models for the aforementioned attributes with Spark. In this talk, I will present how we tackled typical problems that you run into in a production environment such as redundant computation and backfill. For this we are using Spark ML pipelines, User Defined Aggregate Functions (UDAF), and Luigi.

Bio:
Tom joined the user modeling team in Schibsted this year as a data engineer working on predictive models using Spark. Prior to joining Schibsted, he finished his PhD in Utrecht on how conversational agents can use argumentation to understand and to discuss a user’s motivation behind decisions. After his PhD, he worked in a medical care management software company and for Vistaprint working on implementing predictive models for customer lifetime value. In his free time, he plays drums for “The Monkey Men” dances Lindy Hop, and enjoys beautiful Catalunya.