I've been working on a machine learning based forecasting robot. The general objective (can be changed) of the software is to forecast if in the next 1/5/10/15 minutes (something of that magnitude) the current best ask price will change by (F*X)%, or not. where 'X' is the sites fee and 'F'>1 is some factor to ensure I'll also see some profit from the trade.
Now, I know a lot about programming, and I know a lot of machine learning (specifically deep learning), but I do not know a lot about trading strategies. So here are a few questions that I'd be happy for you help.
*There is no need to understand anything about the forecasting method to answer these questions. Imagine you are provided with a black-box (e.g. indicator) that returns predictions upon receiving a market's state, but you can decide what question to ask (focusing on simple yes/no questions)
1) Any recommendation for learning sources? I'd prefer a source with mathematical basis instead of just textual (which usually turn out to be more of a 'philosophical discussion').
2) Do you have any recommendations on a different "target" to forecast instead? If you think the current settings are OK, how to you recommand the robot should react when the (F*X)% raise has reached? exit immediately? wait for it to start dropping? something else?
**Promise to remember(
) these that gave me help that I couldn't have done a thing without it.