Visual Analytics(VA) formally can be termed as a discipline of analytical reasoning aided by interactive visual interfaces that aim at explaining certain hidden patterns in data and algorithms.
I think in today’s data-driven world we have been fairly familiar with how ‘Data Visualization’ can help users derive insights & make better decisions. However, is Visualization & Visual Analytics the same? This is one of the most frequent questions I have faced during my research work at the university or my job. Let us try to understand how do these disciplines relate and where is the boundary of separation.
Information Visualization & Visual Analytics?
Information visualization in simple terms means representing the data in the form of visual metaphors so it is easily interpretable to humans. Generally, the representations are in the form of charts such as Bar, Line, Box plot, Scatter plots, etc. to name a few.
Example: A typical example of where data visualization can be used is during the exploratory data analysis(EDA) process to understand the distribution of features in the dataset. Below, we have a Histogram(built in Qlik Sense) to understand how the feature ‘age’ is distributed. This seems to be quite helpful than manually analyzing data on a textual file. Right?
Visual Analytics(VA) takes the understanding process a step further. What if we could combine visualizations with techniques like Machine Learning, Natural Language Processing, Human-in-the-loop(e.g. interactions) to derive patterns and trends in the data or algorithms? VA enables users to drill down to a problem by leveraging the above-mentioned processes and allows for detailed exploration. Alright, that’s a lot of terminologies.
Example: Here is a Radial visualization(RadViz) I developed using D3.js to understand Spotify’s music dataset features. Each of the points you see in this visualization is a song.
What if I added a functionality to group these songs into two major clusters based on some level of similarities between the songs? Something like this.
Resulting in this.
Now, we see 2 big bubbles(clusters) of songs. So, hypothetically if I like a song from the ‘orange’ cluster, I can look for a similar one from that same cluster and listen to relevant music. Now, that’s where Visual Analytics has played its part! It has allowed me to take advantage of an unsupervised learning algorithm(k-means clustering) in the back-end of the visualization to analyze similar songs in contrast to just understanding the properties of the song through the dataset.
And that brings an end to understanding the relationship & differences between Visualization and Visual Analytics.