Newbie question: Neuronal Networks prediction

Rapido

I don't know if anyone could give some help to start with rapidminer on this problem.
I'm trying to predict some time series of different objects.
For example I have 3 Objects with 3 attributes each on weekly basis (for 2 years).
The goal is to predict attribute u on basis of the attributes of the objects.

My idea was to have a file in this form.

Object Name, weekyear,a1, a2, a3
House1, 012007,u1,a11,a12,a13
House2, 012007,u2,a21,a22,a23
House3, 012007,u3,a31,a32,a33
.
.
.
House1, 012008,u1,a11,a12,a13
House2, 012008,u2,a21,a22,a23
House3, 012008,u2,a31,a32,a33
.
.
.

But I'm not really sure if I it's possible to model time series data with non-time series data.

For example the attributes u1,a11,a12 for House1 are time series and a13 is a non-time serie and so on.


My idea was to transform weekyear with date2Numerical, make the weekyear as ID, normalize the data, apply MultivariatSeriestoWindowExamples and
then use a neuronal net. Do i have to add for every weekyear the non-timeserie data??

I would appreciate it if you could give me a few hints how to start.

Thanks in advance.

Regards,

Mike

land

Hi,
let's start with what you really need in the end to build a prediction model: All input attributes which should be given to the model plus the label must be stored in one example. The attributes which should be used for modeling could very well be mixed: For example if you talk about houses, then the number of rooms is somehow fixed over time, while the house prices in the neighborhood might change over time. While windowing the series of house prices gives you a couple of examples for this single house, you have to include the constant information at every of this example.
I think it would be the best to split the time series data and constant data, build windows from the time series (and thus having multiple examples per object describing on point in time) and then join this and the constant data over the object's id. This should result in the desired dataset for modeling.

Greetings,
  Sebastian