A Student At The Center For Network Science Has Created An Algorithm That Would Help Predict Who Will Die Next In Game Of Thrones

Everybody knows George R. R. Martin is good at killing our favorite characters–be it Ned Stark, Robb Stark, or Khal Drogo. It is almost impossible to know who would die next in this popular HBO show.

Milan Janosov, who is a Ph.D. student at the Center for Network Science at the Central European University, claims to have created an algorithm that can predict the likelihood of a will-be-dying character of this unpredictable show. It may sound insane, but with some graph algorithm that consists of approximately 400 nodes and 3000 edges, he can predict the death of your favorite character.

Milan said that he used scene analysis to create seven different network-based features based on their various dimension of social importance. He called it “the elementary unit of social interaction”. He combined this with the traits of the 61 character that previously died in this show.

Janosov writes — “This problem resembles the well-known churn problem, which can be solved with various classification-based algorithms”.

Support Vector Machine (SVM), is a supervised learning model used for finding the probability of the surviving character that will die before the show ends next year in 2018.

According to this model, Theon Grejoy, popularly known as “Reek”, is least likely to die in the show with the probability of just 0.05. The character who is most likely to die is Tyene Sand, bastard daughter of Prince Oberyn Martel with the probability of 0.95. Daenerys Targaryen who is one of the most favorite characters of the show is second most likely to die with the probability of 0.91. Tyrion Lannister, on the other hand, has a 0.52 chance of dying, according to the model.

As per Janosov, the prediction can be made better by including features such as gender and Nobel clans they belong to. Dr. Guy Yachdav said, “In its daily work, our research group focuses on answering complex biological questions using data mining and machine learning algorithms.” The only difference was that the project was a popular TV show.