Beg need help in importing data

inceptorfull

Hello All,
I have the Excel file to be imported to Rapidminer and I know how to do it, but choosing the arttibutes types make it difficult for me.

the excel contains of :

Year,      company,        sales, ......... some other Ratios,                  and TA( which is the dependent Variable)
2012        1                      222                                                                      500

whtever.

so I want to  know what Right attributes to choose, additionally,

DO I need to Normalize data before putting to Neural network?

MartinLiebig

Do you mean "right" in terms of the best performance in your learner? (e.g. neural net)?

And a neural net usually does not need normalization

inceptorfull

thanks for reply,
No I mean by "Right" the suitable attribute for the company and year variables? will be id, or label or what?
others will be numeric

MartinLiebig

sorry, then i do not understand what you want.

Do you mean roles? Like id, label, and other custom based roles?

inceptorfull

thanks Matrin, Yes that what I meant, since the year and company name will be dummy variables,
so I want after regression or NN or wtvr analysis, I can know for what company in which year

MartinLiebig

Hi again!

The operator you want to go for is "Set Role"

Attributes can have different roles. Every role different from "regular" makes the attribute special. Special attributes are ignored in usual operators. You can specify custom roles by simply typing in a name at the "target role" field of the set role operator.

There are some roles which have a special meaning. For example there are:

Label - The "target variable", so the one the neural net will use to predict it.
id - an id attribute which is needed for some operations.
and some others.

I guess you just want to set the role label to the target variable and all others which should be ignored during learning to some custom role.

Cheers,

Martin

inceptorfull

thanks a lot for your reply, I will try it,

I have other question how to input five years, so I can train only three years and predict the last two?

MartinLiebig

I would highly recommend to use a Cross validation (the operator is named X-Validation). This is way more accurate than a split validation you might perform.

If you really want to use it with a split you can use Filter Examples for this.

If you need help with this, i could post an example process later

inceptorfull

I will really apperciate that, I just dont want to make anything wrong when analyze data since its for Academic purpose, So If you have tutorial or recommended book for that matter will be great, I just donot want to take much from time , in writing example, Thanks again