Technology

Bright Idea: Predictive Analysis

By Drew Atkins November 18, 2014

graphlab_web_0

This article originally appeared in the December 2014 issue of Seattle magazine.

As the big data phenomenon continues its march toward business world ubiquity, many enterprises are compiling as much information as possible on their customers, on their operations and anything else that seems potentially relevant. But what to do with that data once theyre collected? How can they be used to make high-level, effective decisions?

Enter GraphLab Create, a software platform that is changing the way businesses as diverse as Adobe, Cisco, Zillow and Zynga operate. The brainchild of former Carnegie Mellon Professor Carlos Guestrin, GraphLabs goal, he says, is to help data scientists become the heroes of the business world.

Guestrin, who is now the Amazon Professor of Machine Learning at the University of Washington, designed GraphLab to facilitate work on complicated algorithms around predictions and automated deep learning, efficiently structuring raw data while reducing the sky-high computational power this typically demands. If that sounds over the heads of most people, it is meant to be. Guestrin is a big fan of Tableau Software, one of the most popular tools in the data analytics space. But he says that GraphLab is targeting a far more specialized customer.

GraphLab targets the people who can write code, the data scientists, and helps them do what they do better, Guestrin explains. A single data scientist, for example, can use GraphLab to do what it used to take several scientists to do. By contrast, he says,Tableau is targeting the business analysts, the people who can use Excel or less. Python [a programming language] is a starting point for us.

Further, where Tableau teases out patterns in current data, GraphLab concentrates on where the data are going. This predictive focus, he says, is the reason 200-plus companies have been using GraphLab just in beta. Zillow uses it for improved housing market predictions. Online game maker Zynga had a large cluster of computers dedicated to crunching data around user recommendations; GraphLabs program brought that need down to two computers.

Whether its Netflix recommending your next movie or a city determining how motorists navigate the roadways, an enormous amount of data analytics is involved. Tableau has grown to a multibillion-dollar market cap helping businesses understand and visualize information. GraphLab could be poised to make real waves as well, helping businesses effectively capitalize on the big data revolution and, in Guestrins words, get something out of their data that transcends it.

Follow Us