I have done this type of project before, seems like a problem of supervised learning, we can use Random Forest, Support Vector, K nearest neighbour and Naive Baye's Classifier to check which has a better prediction precision, thereby using it and plotting the roc curve. Hoping to hear from you soon.