 A few videos ago, we already peaked at the Human Development Index data. So now, let's see if clustering the data presents any new, interesting findings. In this video, I will attempt to visualize these clusters on the GeoMap. Orange comes with a Geo add-on, which I will need to install first. Add-ons can be found in the menu under Options. Here, I select the checkbox for the Geo add-on. The add-on will install, and orange will have to restart. As you can see, I now have another group of Geo widgets. OK, back to my Human Development Index data. I can find it in the Datasets widget by typing HDI in the filter box. Double-click to load the data. Whenever I load a new dataset, before doing anything else, I like to examine it in the data table. This data includes 188 countries that are profiled by 52 socioeconomic features. To find interesting groups of countries, I first compute the distances between country pairs. I will use Euclidean Distance, and make sure that the features are normalized. As this dataset includes features with differing value domains, normalization is required to find a common ground. In hierarchical clustering, I use the word index and label the branches with the names of the countries. Word index joins the clusters so that the data variance in the resulting cluster is decreased, and often produces dendrograms with more exposed clusters. Our clustering looks quite interesting. We see some clusters of African countries like Swaziland, Lesotho, and Botswana, and we also find Iceland, Norway, and Sweden group together, and some countries of the ex-Soviet Union. By zooming out of the dendrogram, I can find such a cut that will group the countries into four different clusters. So let's take a look at them on the map. To do this, I first need to add the geocoding widget to the output of hierarchical clustering. The widget is already set correctly to use the country name as an identifier. By opening the data table, I can double-check the result and find that every row now includes extra information on the cluster. This comes from the hierarchical clustering widget and also information on the latitude and longitude of the country added by geocoding. I will now pass this data onto the choropleth map and just quickly change the attribute for visualization to cluster and the aggregation to mode. Well, here it is, the countries of our world clustered. Looks interesting. Well, we refer to as developed countries are in one cluster. Countries in South America are clustered with some countries in Asia and we also find a band of Sub-Saharan African countries. It's fascinating that although the data did not explicitly encode any geolocation information, the clusters we found were related to the country's geographical location. Our world really is split into regions and the differences between countries are substantial. Did I say differences? I really did an abysmal job of explaining what exactly these clusters are. That is, what are the socioeconomic indicators that are specific to a particular cluster? Are some indicators more characteristic of clusters than others? Which indicators characterize clusters the best? So many questions. It is time to reveal the trick of cluster explanation, but that I'll leave for our next video.