What is the purpose of learning data visualization for machine learning?

Data visualization is used to convey information (e.g result or insight from machine learning). You wouldn’t want your finding to ends up with you. Knowing the correct principle of data visualization enables you to convey correct information that is easy to be understood, less explanation needed and does not pain the eye. It is also used as aid tools to explain your findings to non-IT person and also stakeholders.

If you want to practice machine learning, you should learn data visualization because you have a superb pattern matcher in your head. The human eye is wonderful at extracting patterns from images.

Here’s a classic example: Anscombe’s quartet – here are four datasets, having an equal mean, standard deviation, range, correlation between x and y, linear regression lines, and yet are very different from each other.

By looking at the scatterplot, you immediately get the difference between these datasets, while if you only saw their summary statistics they would seem identical.

This is an example of the power of exploratory data visualization, a powerful tool in the data scientist’s kit.

There’s also expository data visualization, through which you can communicate your findings to other people. But IMO this is a secondary use case: Exploratory data visualization enables you to get a grip on a new dataset or a specific subset, including central tendency, outliers, general distribution and whatever does not feel right with the data. And believe me, many times something with the data does not feel right.