Graph Algorithms: from Theory to Useful with PlaceIQ
Details
Got a great talk this month with Jeffrey Picard from PlaceIQ.
Abstract:
At PlaceIQ, massive amounts of location and movement data are used to generate insights into consumer behavior, and many of the problems that arise naturally lend themselves to graph-based solutions.
We will explore two separate cases where graph algorithms proved incredibly useful: conflating device IDs across disparate mobile ad data and understanding the value of driving routes most frequently used by target audiences. Often times, as we will see with the case of conflating device ids, the problems can be solved with off-the-shelf algorithms like connected components, and challenges mainly involve achieving scalability or dealing with noisy data. Other times, as with the driving route analysis, custom algorithms must be developed that leverage specific attributes of our data and produce usable results.
As we walk through these real-world cases, we will also cover topics in elementary graph theory, survey various graph libraries, and touch upon some issues that may arise during implementation.
Bio:
Jeffrey Picard is a Software Engineer at PlaceIQ, working on developing production pipelines and analyzing data for fun and profit. Originally from New Hampshire, he has a BS in Math from the University of New Hampshire and is pursuing an MS in Computer Science at Columbia. Despite the prevalence of Java in the big data ecosystem (or because of it), his favorite language remains C. Long live pointer arithmetic!
--
A giant thank you to Spotify for hosting & providing the pizza and beer.