How realistic is it to model a forecasting system based on the interpretation of historical facts?

Question

big data

If we take an index of world events (for example, parse from news archives), decompose events as vectors (I have little idea what I mean here, here you can play with the parameter and set the Nth number of dimensions and decompose events into these vectors), train the system on segment 1990-2010 and force to predict for 2016 (we are talking about the dollar or the NYSE). Maybe there is something similar?

Answer 1

It's not real at all. Neither the dollar exchange rate, nor the events on the stock exchange, nor any other events in the real world can be predicted on the basis of statistical data. There have been attempts and they have failed.

Answer 2

And why "pervert" with the digital system, if in this case the analog one is perfect, that is, the brain of a person who has worked enough in this direction?
Not enough brain RAM to store a large amount of facts? No problem, connect an external media like "notepad" through the "hands" interface and save.
Take an example from Chinese engineers, why there are so few robots in factories in China, because they could definitely afford much more, but the best robot is a person.
It is not difficult to study history as a science, drawing conclusions and learning to predict what really lends itself to it, for example, wars and uprisings do not lead to anything good.
I understand that you want to dream, but they don’t choose the subject of your dreams, or is it difficult to make a choice from study and there, but maybe it’s better to think about automating what really needs it, rather than doing nonsense?

Answer 3

only in the Matrix