People talk about an ever-changing “television landscape,” but what does it actually look like? And if it is in fact evolving, then why do we use push shows into the same old genres like sitcom and drama? TV shows may have followed a simple formula once upon a time, but now with more diverse shows than ever it’s time to reevaluate things. And instead of arbitrarily assigning new labels, we will let the data tell the story.
Conveniently for you, dear reader, our database of show attributes allows us to calculate similarity scores for a given show to every other show. Using this, I was able to build a data set of show similarities, and represent it as a force-directed graph. Each edge in the graph represents a strong similarity between two shows. Since these have a similarity score, they are weighted accordingly and the graph will naturally cluster the most alike shows together. A simple graph partitioning algorithm was applied to color the graph into 20 clusters for the purposes of this visualization. This coloring, and these clusters, help illuminate what the true modern genres of television are. There may not be an agreed upon name for each group, but they certainly start to reveal themselves through this visualization. This quick clustering serves well as an initial step, but more work will have to be done to find and understand the nuances within some of these larger groups.
So for now, go ahead and explore the graph to help get a feel for the otherwise nebulous television landscape. Click the picture of the graph below to fire up the interactive visualization. Mousing over a node will pop up the show poster, and clicking and dragging can help you investigate the sometimes tightly clustered data. Enjoy, and keep an eye out for more posts analyzing this data in the future.