I’m looking to develop a Machine Learning/Deep Learning model to predict whether a stocks current day data is noise or is a signal for Over/Under performance vs a benchmark for the next day. I would like to get Over/Under prediction and a probability of the signal. Should I use LTSM RNN even if the model only takes in todays data to predict tomorrow or is Hidden Markov Model the way to go?

Looking for advice on direction and I am in data gathering phase.