I have a tabular dataset of financial transactions with a target binary variable 'isFraud' which indicates if the transaction is fraud or not.
I want to build a model that given some past trasactions, will output one or more transactions that are likely to be non-fraud. I thought of predicting some future transaction (=rows) based on the past, and the predictor will learn only from non-fraud transactions.
My question are the following:
- Is there a way to predict future rows of a tabular dataset based on existing ones?
- Do you find this way (of predicting non-fraud) plausible?