A very important area nowadays is the prediction of fraudulent transactions in banking. When dealing with this data, the problem we run into is imbalance, the transactions available are mostly legitimate and only a small percentage fraudulent. It is therefore very difficult to train the models with this type of data.