Virtual reality is mainly a form of entertainment that provides immersive experiences such as sitting courtside at an NBA game, shooting aliens with your hands, starring in a horror movie and more. However, there has also been a lot of speculation about the potential applications of[VR in the big data world. In industries where immersion is required, such as remote surgeries, trainings across industries such as auto repair and military, virtual reality provides major value. Will there be any applications for virtual reality in data analytics and visualization though?
Being the gaming and tech nerds we are at TCB Analytics , we purchased an Oculus Rift with two touch controllers (Figure 1), three touch sensors and decided to try it out for ourselves.
A setup with three sensors instead of two allows for full room-tracking similar to the HTC Vive’s room-scale capabilities at launch. This means that no matter where we move in the room, or if we turn around 360 degrees, the sensors will be able to track our movement.
The setup consists of two sensors facing front and one sensor diagonally to the right back wall (Figure 2).
The only data visualization application currently available is a free demo called datavizVR and it is available on the Steam Store.
The demo comes with a few pre-loaded datasets that you can visualize. We selected a dataset about video games, which consists of the following variables:
- Global Sales
The user then chooses variables to assign to the X-axis, Y-axis, Z-axis as well as which variables should indicate the color, size and label of each point (Figure 3).
What kind of questions would someone want to answer about this kind of dataset? A few that come to mind are:
- Does Rank correlate positively with Global Sales?
- Do specific Genres of games sell more than others?
In order to answer the first question, we assigned Rank to be the size of the point and color to indicate Global sales. Year, Genre and Game were put on the Z, Y and X axes, respectively.
It became easy to see that smaller circles (bottom right, small blue circle) had poor sales (Figure 4). One can then point to a specific circle with the touch controller, click on it to see more information such as Genre, specific Game, or Publisher. You can also rotate the cube of data, step inside it and interact with every single point individually (Figure 5).
The question now becomes: “Is interacting with data in this manner more valuable than 2D or static charts?” The three axes are overkill for answering our initial question about Game Rank and Global Sales.
How many executives or analysts will don a VR headset at a meeting, whether in-person or a teleconference?
Is data visualization in VR nothing more than hype right now? We have barely scraped the surface with interactive 2D plots and ensuring that insights are clearly communicated to business stakeholders.
Not only that, but interactive 3D charts have been around for several years now. For example, take a look at 3D protein structures via the NCBI. A user can manipulate the 3D representation of a protein, turn it, rotate it, and zoom in on it (Figure 6). These are the same capabilities that we have in VR, only using a mouse instead of our own hands.
There is also NASA’s Many Eyes application, which enables users to fly along with NASA on space missions or view Earth from a global 3D perspective.
As data practitioners, we should remain innovative and think creatively about the potential applications of new technologies to our industry. At the same time, let’s also retain our skepticism and be realistic about providing business value.
In its current state, standalone data visualization in VR provides nothing more than a cool Minority Report experience by enabling us to interact with and move data around with our hands. However, we do believe VR will have promising applications in various data-related industries.
For example, in remote surgeries we can imagine the surgeon looking to a virtual clipboard that shows the patient’s vital statistics, or a dashboard conveying predictions about potential adverse events in real-time. In the same way that data scientists currently collect, analyze and visualize data, we will need to be thoughtful about how we embed those visualizations in a virtual world. We will need to identify the right times during those experiences to provide haptic responses such as buzzing a controller if the surgeon is too close to a specific organ with an instrument. We will have to adjust to a whole new world of assumptions and design constraints for virtual reality experiences.
The datavizVR application can certainly be improved upon, hopefully allowing users to upload their own datasets, providing legends, and the ability to export images of 3D visualizations, but for the time being it remains a shiny demo.
Data scientists will eventually incorporate virtual reality into their workflow depending on the use case and its relevancy to their domain. Just don’t expect C-level executives to want to receive Excel reports, KPI dashboards or bar charts in VR any time soon.
What are your thoughts on how data scientists will use virtual reality? Leave a comment below.