feature rank manually

fatimidveil

hii every one , i am working on my thesis .i collect data set from public and private sector colleges ,my data set consist of 50 variables .. some variables are for general purposes and it does not effect on my result , i want to generate but when i apply feature selection algorithms they select those variables as high rank .
can i manually rank the variables that are highly effect my result ? are i want to remove those general variables .
thanks waiting for your suggestion .
regards.

Telcontar120

You can also simply use "Select Attributes" and only pass through attributes that you want to consider for modeling inputs into any machine learning operator.   This will make sure that any items that are not appropriate don't end up in a model.  You can join the dataset back to a copy of itself to re-add those attributes later if you want them for reporting purposes.

varunm1

Hello @fatimidveil

Did you check the "Weight by User Specifications" operator? This helps us assign weights for attributes (Variables).

https://docs.rapidminer.com/latest/studio/operators/modeling/feature_weights/weight_by_user_specification.html

 i want to generate but when i apply feature selection algorithms they select those variables as high rank .

Yep, this can happen if they are statistically significant in predicting the output labels.

Hope this helps.

Telcontar120

You can also simply use "Select Attributes" and only pass through attributes that you want to consider for modeling inputs into any machine learning operator.   This will make sure that any items that are not appropriate don't end up in a model.  You can join the dataset back to a copy of itself to re-add those attributes later if you want them for reporting purposes.