10 things about Tableau 10: Clustering

13.09.2016

10 things about Tableau 10: Clustering

Continuing on with my “10 things about Tableau 10” blog series, it’s time to dive into another new feature! Following up on my previous post about the interface, this time we’ll be looking at Tableau’s new Clustering tool.

Clustering

Clustering, specifically k-means clustering, is probably my new favourite feature in Tableau 10. k-means clustering is a method of vector quantization, which groups data points together by the nearest mean value. With the new drag and drop tool, Tableau automatically finds patterns in your data by using the algorithm. You can use clustering on any type of visualization, from scatter plots to text tables and even maps!

Previously, clustering was available in Tableau through the R library extension, which required the user to create a calculated field calling an R-script. Even though this wasn’t terribly difficult, it nevertheless required an instance of RServe to be present. The new built-in functionality makes creating clusters a snap!

clustering-tableau10-example

To use the clustering tool, just access the Analytics tab, and drag the Cluster model to the canvas. Tableau uses the measures on the canvas to create a set amount of clusters. From the Cluster control panel, you can define what variables (measures) to use in the cluster analysis as well as manually define how many clusters Tableau creates. To add more measures, just drag them into the box!

clustering-dynamic-example

Clusters can be renamed and set colours manually, much in the same way as you would for colour legends or dimensions. They can also be used together with other visualizations on a dashboard to highlight the values in clusters. The number of clusters will also change dynamically if not specified separately.

As with the Forecasting tool, it’s important to understand how the underlying model has calculated the results. By right clicking on Clusters mark and selecting the Describe clusters menu option, we can dive deeper into the mathematics of the k-means clustering. Tableau will show you the number of datapoints, the sums of the centers as well as the variables of the model. The Models tab goes into futher detail about F-statistic and p-values.

tableau10-clustering-describe

 

To download a free 14 day trial of Tableau 10, go to our trial download page and discover Tableau 10 for yourself!

Share
Contact Person

Blog writer

Karri Linnoinen

Bilot Alumni