Hi guys. Since there seems to be an ML component, I was wondering how this was built. Do you start with a basic hypothesis and see if it holds for some training

Hi guys. Since there seems to be an ML component, I was wondering how this was built. Do you start with a basic hypothesis and see if it holds for some training set and then refine it or right off the bat go with a generic classifier like SVM (with some features in mind) or something like that and then tweak it? Also, how did you choose the dataset for this ? This project is very cool. Congratulations and good luck!