I wanna know if the row entries are actual ticks or not, and whether the targets are the values for this particular time, or is it the resulting target after a tick with the given features?
Assuming the features are a function of time and the target is also a function of time, is the data related as follows?
Features(t) → target(t)
Features(t) → target(t+1)
You can’t really do time-series with this data. You can’t match up rows from different periods. The target is 4 weeks (20 trading days) in the future relative to the features – that’s what you are predicting.
But are the rows chronologically ordered?
The eras are chronologically ordered. The rows within each era are of the same time period. Each row is a different stock, so there is no time ordering within an era – an era is a snapshot of time of X number of stocks (X = number of rows in era). The row order within an era (i.e. the order of the stocks) should be considered random. And as I said, there is no way to link up a specific stock row in one era to the same stock in a different era (and it might not even be there anyway – the exact selection of stocks isn’t the same in each era either.)
Alright, thank you for the explanation.