Altair RISE
A program to recognize and reward our most engaged community members
Nominate Yourself Now!
Home
Discussions
Community Q&A
When AutoModel asks you to "Deselect" red or yellow attributes, is AutoModel doing feature selection
tonyboy9
Find more posts tagged with
AI Studio
Accepted answers
RolandJones
Hi
@tonyboy9
,
The colouring of different attributes is based off their measured quality, the results of which you can see on the right hand side columns. For example, the top 2 columns appear to be marked yellow given their zero correlation with your target. When AutoModel is asking you to deselect, it's asking you if you'd like to remove lower quality features based off the measures. Of course, there will be instances where you want to keep certain features anyway, and other times when it would be inappropriate to keep features labelled green/good.
Hope this helps,
Best,
Roland
RolandJones
Hi
@tonyboy9
,
It comes down to how in depth you want to improve the input data to your model to improve predictive power. The AutoModel stage which you showed is a set of statistical measures where we grade each variable and then select or deselect based on quality - this is a first step of feature selection. What you'll find is that later on when you use AutoModel, you get the opportunity to do advanced Feature Selection and Feature generation. This would be termed as a topic "Feature Engineering". I'd recommend this video as a nice starting point on why we do this:
https://academy.rapidminer.com/learn/video/feature-engineering-intro
Within RapidMiner Studio, there's a number of different ways we can approach it, in order to produce the best outcome.
Best,
Roland
All comments
RolandJones
Hi
@tonyboy9
,
The colouring of different attributes is based off their measured quality, the results of which you can see on the right hand side columns. For example, the top 2 columns appear to be marked yellow given their zero correlation with your target. When AutoModel is asking you to deselect, it's asking you if you'd like to remove lower quality features based off the measures. Of course, there will be instances where you want to keep certain features anyway, and other times when it would be inappropriate to keep features labelled green/good.
Hope this helps,
Best,
Roland
tonyboy9
Okay, Roland, thanks for that. Perhaps I can take this one step further. What good are all the feature selection operators available to build a process? I tried to follow a response explaining the roles of these operators, which just reads overly complicated. Why not just rely on AutoModel decisions?
RolandJones
Hi
@tonyboy9
,
It comes down to how in depth you want to improve the input data to your model to improve predictive power. The AutoModel stage which you showed is a set of statistical measures where we grade each variable and then select or deselect based on quality - this is a first step of feature selection. What you'll find is that later on when you use AutoModel, you get the opportunity to do advanced Feature Selection and Feature generation. This would be termed as a topic "Feature Engineering". I'd recommend this video as a nice starting point on why we do this:
https://academy.rapidminer.com/learn/video/feature-engineering-intro
Within RapidMiner Studio, there's a number of different ways we can approach it, in order to produce the best outcome.
Best,
Roland
Quick Links
All Categories
Recent Discussions
Activity
Unanswered
日本語 (Japanese)
한국어(Korean)
Groups
