Is it possible to use create a machine learning model to determine the best arbitrage opportunity

this might be a bit of an open question but I would just like to understand further on the subject as I am rather new to machine learning and arbitrage trading.

Would it be possible to develop a machine learning model to determine which arbitrage trading strategy would be best, in this case, I am focusing on simple exchange arbitrage (e,g. the immediate buying and selling of an asset) and triangular arbitrage.