Episode 30 - Entropy is a Decision Tree: Game of Thrones Predictions
Some data scientists are attempting to use data and machine learning to predict the outcome of the next season of game of thrones. In this solo show I discuss the merits of this approach, give an overview of decision trees, adversarial problems, and the role of information theory.
Articles
Predicting Death in Game of Thrones on the DataRobot Blog
Check out the Kaggle Competition for the Game of Thrones Dataset
Epic Rap Battles of History, Martin vs Tolkien whose lyrics suggest different entropy levels in Westeros vs Middle Earth.
Previous Episodes Mentioned
Episode 10: Zuckerberg Testifies - the difficulties in adversarial machine learning problems
Episode 19: Information Theory - the information content of fiction
Episode 23: Natural Language Processing - how this could be used to analyze writing and create these datasets
Episode 24: Reinforcement Learning - examples of problems that would be exceedingly difficult to do with handwritten rules
Episode 27: Fat Tails - On the nature of unpredictable disasters like that which can occur to certain characters on Game of Thrones, and Thanksgiving Turkeys