The Data Mining Forum 
IMPORTANT: This is the old Data Mining forum.
I keep it online so that you can read the old messages.
Please post your new messages in the new forum: https://forum2.philippe-fournier-viger.com/index.php
Beginner Question About Building Groups/Features
Posted by:
JackDaniels
Date: December 19, 2014 09:55AM
Hi Guys,
i'm new to the topic of data mining. I want to predict the outcoming of a variable. For this task, i want to make the most out of the given data. Some of the discrete data is very useful but they come with a large number of parameter values/levels(?). So i want to group them or make them continuous. My question: Can i make new groups/values based on the outcome-probability of the prediction-variable? Does this make sence or do i break some mathematical law of data mining?
THX
i'm new to the topic of data mining. I want to predict the outcoming of a variable. For this task, i want to make the most out of the given data. Some of the discrete data is very useful but they come with a large number of parameter values/levels(?). So i want to group them or make them continuous. My question: Can i make new groups/values based on the outcome-probability of the prediction-variable? Does this make sence or do i break some mathematical law of data mining?
THX
Re: Beginner Question About Building Groups/Features
Posted by:
webmasterphilfv
Date: December 19, 2014 12:21PM
I'm not 100 % percent sure that I understand the question. I think that what you want to make some prediction based on an attribute and then use this prediction to group other attributes.
In my opinion, maybe that it would be ok. But there might be some risks to increase error.
If you have too many attributes and wants to reduce the number of attributes, there exist some algorithm for dimensionality reduction such as PCA. I'm not sure if it is what you need but you may have a look.
In my opinion, maybe that it would be ok. But there might be some risks to increase error.
If you have too many attributes and wants to reduce the number of attributes, there exist some algorithm for dimensionality reduction such as PCA. I'm not sure if it is what you need but you may have a look.