Five trends to watch in visualization

  • BLOG
  • July 30th, 2019

Some still believe that the traditional flavors of  big data analytics – descriptive, inquisitive, predictive, and prescriptive – represent a linear evolution. But the real democratization of analytics will occur when we answer the “What” in a way that enables businesses to better answer the “Why”, “How”, and “What’s Next?” And that will require the right approach to visualization.

Visualization – the use of images, diagrams, and animation to communicate information – has transformed from its earliest avatar of medieval cave paintings, but its purpose has largely remained unchanged: To depict past events in a comprehensive and aesthetically pleasing manner.

Until now. Visuals now have a growing role in helping to predict and identify hidden patterns within data. Here are 5 visualization trends with the power to transform how we think of advanced analytics:

  1. Topological Data Analysis (TDA.) With TDA, complex multi-dimensional data is normalized and plotted in an n-dimensional space forming a ‘data cloud’. Look for applications in sports (player movements), wearables (improve customer experience), e-commerce (discover customer preference), retail (better assortment planning), decision hierarchy, fraud detection, search (combine social data with search results), and healthcare (prevent diseases by studying patterns.)
  2. Voronoi Tessellations. These visuals use a classification scheme to divide a problem space into regions. Analysts set points, known as seeds, on the diagram and for each seed there exists a corresponding region consisting of all points closer to that seed than to any other. These have applications in epidemiology, polymer physics, computational chemistry, cancer diagnosis, robot navigation, location selection, modeling retail trade areas, and demand forecasting.
  3. Multi-dimensional projections. 2D/3D data projections help reduce the dimensionality of multi-dimensional data and project into lower dimensions to allow for easier comprehension. They have wide applications in areas like time-series analysis, understanding correlations, optical character recognition, signal processing and fraud detection.
  4. Spanning trees. This is a networking technique to resolve common runtime and optimization issues faced in conventional K-means clustering, and can be used in all network optimization problems like route-optimization, cabling, electrical, hydraulics, etc.  Spanning trees are also used to measure correlations between different stocks, product affinities in retail, and the like.
  5. Community detection algorithms.  These help identify community groups in social networks based on demographics and behaviors. Metabolic networks also have communities based on functional groupings, being used to study bird migration, social networking problems, gene pools, and the spread of disease. You could also use the same technique in detecting communities of physicians in the pharma industry, or customer segmentation based on online behavior.

These are just a starting point for the conversation. Visualization holds real promise of providing a query-free and agile approach to data analytics services.