If there are 5 ducks flying in different directions that does not give any information. But if they join in flying south and they make turns in the very same moment you know autumn is coming. Algorithmically speaking this is pattern recognition and it is very difficult to program it. Though human brains can spot patterns relatively easily. The brain can be helped to be more efficient in this task by computer programs.
Assuming your timeframe is compatible, I would look at the CFTC's COTS report. Google COTS trading systems for specific ideas, but you basically look to piggy back on or fade certain categories of traders.
keep longer term timeframe eur/usd and dx on your screen look for turns on major supp;res and use market profile and 5 and 1q5 and 30 minute tl's for scalping
I am good at programming pattern recognition systems. I used to do facial recognition and speech recognition. But how to use pattern recognition techniques here in trading?
I downloaded some CFTC data and throw into my neural networks to do the forecasting but didn't get anything meaningful. Maybe I am not using the data in the right way...