The two Quantitative sides

In one of my last post, I said that I got microstructure data from IEX exchange. Know There are two solution for a quantitative analyst in front of this kind of data :

Take all the data and do Deep (blind) Learning and by analysing severals combinations of features and windows of observation by making principal component analysis hoping to find a high performing model. I built a program for this kind of work named snake analysis which discovered profitable models with daily data but this kind of system use lot of computing amount and is not best fitted for a single independent searcher like me.

The second choice which is mine is to apply known theoretical models mathematically and statistically tested and convert them into algorithms and confront them with real microstructure data coming from the markets. This choices require great understanding of markets, orders types, …. But it’s the most disciplined way of working with big data of exchange tapes. Who said that mathematics are not useful. For sure he want and like money like me.