Algorithm predicts which Game of Thrones character will die next

An automated machine learning platform can predict who is most likely to die next on Game of Thrones using an algorithm.

From creating code names to creating scripts with fake endings, screenwriters take different measures to keep storylines a secret. To protect their final season, TV show ‘Game of Thrones’ decided it would no longer give scripts to actors. It is increasing its security around storylines as the show continues to grow in popularity. Actors are receiving their lines in an unusual way – through ear pieces. However, DataRobot, an automated machine learning platform has found a way to predict which character on Game of Thrones will die next.

Using machine learning and a new algorithm called LightGBM, the company analysed 2,000 characters from the show. The analysis covered various characteristics including the house each character belong to, any relatives who previously died, gender, age and their nobility status. DataRobot automatically builds a blueprint for each specific algorithm, in order to optimize its data. The platform uses a variety of techniques and methods to gain insight into the information it collects. For example, feature impact is a technique to see which characteristics entered into the system is the most influential. DataRobot then uses a method called partial dependence to investigate the relationship between each character feature and death.

Nevertheless, the algorithm can only make predictions about the future of the show based on past data. It cannot take into account any elements of surprise or plot twists by the writers. Here at Springwise, we have published many innovative uses of algorithms. In Russia, computer scientists developed an algorithm uses data from Instagram to find travel spots preferred by locals. A new software from New Zealand uses algorithms to help doctors diagnose patients. What other industries could adopt DataRobot’s automated machine learning platform for problem solving?