Thursday, September 26, 2013

UI for Big Data Visualization

     Humans are not very good at numbers. When we see a lot of numbers and data, we tend to lose interest and even get dizzy. When we are dealing with big data, we are moving from a lot of data to innumerable data. How do we draw an insight out of those virtually incomprehensible data. In UX Week 2012, Jonathan Stray offers an interesting point of view for data visualization.

Human Perception

     Human can process up to 5 objects without thinking. We also have 12 channels for visual perception, e.g., color, pattern recognition. This doesn't mean that we should present 5 objects using 12 channels to the audience at once. It does, however, specify our limit for perception.

Make it simple

      Do not put greater than 5 things on one slide. Three would be the optimal value. Make it simple. People will not be stress and will be likely to be engaged with us.

What Attracts the Audience     

  1. Cluster: People are interested in seeing groups of things. Groups make a large data set much easier to understand.
  2. Outlier: People are interested in something that stand out from the group. These can be based on location, color, orientation, or other attributes.
  3. Extent: People are interested in the amount like the range of salary.
  4. Correlation: People are interested in trend that can be concluded from a data set. Correlation breaks down the original information (i.e., data set) into a few interesting points (e.g., linear increase in population)


     Unfortunately, we usually have a lot of information, usually more than 5 points. Especially, big data has various interesting dimension to consider. This is where visualization comes in handy.
Visualization is about interaction. It shows few information to readers at a time. It can direct users' attention to other pieces of information, when needed.
"A visualization turns something you want to find to something you can see."

Directing Attention

     Here are three effective ways to direct attentions:

  1. Filter: Show only what's important
  2. Category: Make categories. Computer can help by we need to define display metric and clustering algorithm.
  3. Annotation: Put related information onto what you'd like to show. These can be notes, tags, snapshot, history, and so on.

No comments:

Post a Comment