Who dies in Game of Thrones Season 7 and why is it relevant to your business?
In this blog I’m going to set out how I’ve used data science to predict the deaths of characters in the upcoming Game of Thrones TV series. I can hand on my heart confirm that I’ve no insider knowledge of these facts and by using a very simple model, my predictions are that these characters WILL DIE:
- Daario Naharis
- Gregor Clegane
- Meera Reed
- Melisandre
- Tormund Giantsbane
- Podrick Payne
So although I’ve had fun doing this, you may also ask, how is this relevant to data science in my business?
Simply put, data science is about using scientific thinking and computing power to make the best use of your data. There are many different approaches, depending on what you need. In this Game of Thrones example I’ve used machine learning techniques, and make predictions about which characters will die based on a very simple classification model. Despite its simplicity, it achieves a reasonable degree of accuracy; in this case 75%. It is based on the concept of supervised learning. It takes a set of example data, each entry having a category attached to it. We use the data set to infer, or learn, rules which will let us categorise other data.
This is just one technique a data scientist may use and here’s an example of how it could be used in a business related problem. If a grocery shop had demographic information about some of its customers, it could use their shopping behaviour to infer demographics about other customers.
Some of the rules you could use for this would be very simple: how can you tell which customers own cats? They buy cat food.
What about other categories? There may not be simple rules in those cases, but maybe we can devise a complex series of rules, such as a decision tree, that allows us to categorise other shoppers.
For example if a customer buys at least 2 pints of milk, buys a particular brand of sliced bread, and so forth, then we categorise them as having teenage sons. (I used a more sophisticated version of this sort of model in the Game of Thrones predictions.) A human would struggle to go through the data and discover those rules, but it is the perfect job for a computer.
If you want to find out more about the mechanics of this, I’ve written up another blog that sets out the technical aspects.
