Multivariate Regression in reservoirpy about reservoirpy HOT 5 CLOSED

Hi @mustafakucuk0,
If you want to predict a Target column, you probably want to use it as output only. Thus you would only have a 4 dimensional input matrix with "Open, Close, High, Low".

from reservoirpy.

Hello again, I am everywhere hahaha. First of all, thank you for your kind answers. I appreciate that. My theoretical part in multivariate regression may not be enough, but I understand the main aim of the theory. In the above case, I want to predict target values using other features. So, using "Open,Close,High and Low" I want to forecast Target value of 1 step ahead. Is that logical? I'll try to elaborate on it with structured format. I am making the numbers up, so do not try to find a connection.

    Open Close  High    Low  Target

01 15 20 20 15 2
02 18 22 22 18 3
*
*
/ a b c d n

So, my data looks like this; is there possibility to forecast the target value of second row by using 4 features of first row ? When people try to work with multivariate time series, they include the Target value as a feature, but I think it would drive the model to overfitting, but I am not sure about it either. That is why I asked my question with including target value as a feature.

Thank your for time and this library again.

from reservoirpy.

Hello again,
Ok, now I understand better.
Yes you can try to predict the next step. In this case you should not include the target in your input, otherwise the model would probably do not learn the correct thing.

So you just need to do:
Open, Close, High, Low ---> reservoir ----> Open, Close, High, Low

Thus, the input and output dimensions are both 4, you predict the same variables you give in inputs.

Does it answer your questions?

from reservoirpy.

Hi again. I understand you suggestion but what about the time when we have Target value? Using these 4 features, I want to make the model predict the next Target value.

from reservoirpy.

For the time you just need to consider the next time step if you want to predict the next values,
or more generally the 'delta' time steps ahead:
[Open, Close, High, Low] (t) ---> reservoir ----> [Open, Close, High, Low](t + delta)

from reservoirpy.