This project uses Zachary's karate club, a graph representing the social network of a karate club that split into two groups after a conflict between the owner and the karate instructor. The goal is to embed the nodes into 2-dimensional vectors using Node2vec embedding and then assigning each member to one of the two groups formed after the conflict using Kmeans clustering.
The model performed well: all members except one were correctly assigned to their group. The member that was incorrectly assigned is member #9. After spending some time hyperparameter tuning without avail, I did some research and found out that member #9 was also the only member incorrectly assigned by Wayne W. Zachary in his club assignment. We can therefore safely assume that the error is not due to the embedding method or tuning process but rather information that isn't represented in the graph structure and edge weights.