Visualizing High Dimensional Data In Augmented Reality

Benjamin Resnick
Jul 3, 2017 · 6 min read
Creative Commons license

Imagine walking into your office on a Monday morning, just a couple years from now. You pour yourself a cup of coffee, check the news, and then put on a pair of AR glasses. You find yourself surrounded by a sea of gently glowing, colored orbs. The orbs represent all of the data that drives your business. You know this data well. The patterns and colors of these orbs are like a fingerprint. But there’s something atypical about the data floating over the coffee maker. You reach out and select that data. A summary of all the relevant details appears on a nearby computer screen.

If something matters to your business, your systems track it. When you want to consume all of that info, you use this immersive visualization; bursting beyond the bounds of a computer screen, information dense, efficient, and aesthetically pleasing.

My team at IBM is working to make the experience described above real. Immersive Insights is an augmented reality data visualization app. Check out our progress in the video below:


This article outlines a technique that my team has been working on recently.

Matrix of 2D scatterplots, juxtaposed with Immersive Insights
  1. Visualize the data using IBM Immersive Insights
  2. Iteratively label and color-code the data according to an evolving understanding of embedded relationships.

Instacart Analysis Code

How did we create the visualization shown in the attached video?

Results of the Analysis

Users who didn’t purchase any organic foods were tightly clustered within the latent space. This finding was a compelling piece of evidence which supported a qualitative observation drawn from the visualization: much of the variance in Instacart purchasing patterns appears to be between users who purchase premium items, and users who prefer lower cost versions of similar items. This difference between cost conscientious and premium purchasers has meaningful implications for Instacart’s marketing, promotional, and recommendation strategies.


This article presented a technique for analyzing Big Data using AR. This technique is most applicable to Data Scientists who are preparing to create Machine Learning models.

Helpful Links

Inside Machine learning

Deep-dive articles about machine learning and data.