Altair RISE
A program to recognize and reward our most engaged community members
Nominate Yourself Now!
Home
Discussions
Community Q&A
Optional use of variables with prediction Automodell
MikeGer
Sorry for the newbie questions, I’m very new to this but am trying to learn!
I created a csv dataset with around 20 variables (columns) which should be able to explain one other variable (column).
After setting everything up and choosing the function „prediction“ with the automodel feature, which btw is a blessing, I found myself having 3 main questions:
1. some of the values I have have a
2. Should o use the feature „classification“ when setting up the
3. Do the models use all variables in order to explain the dependent one, or do the models choose between the ones that help and don’t help the prediction?
Sorry for the questions, I tried to google but couldn’t find answers...
thank you in advance and best greetings from Germany!
Find more posts tagged with
AI Studio
Accepted answers
lionelderkrikor
Hi
@MikeGer
,
I will answer to the question 3. because your questions 1&2 are incomplete (please complete these questions so that we can understand them...)
The short answer to your question 3. is "not necessarily".
First step, AutoModel select only the relevant feature based on the "quality" of these features (correlation to the target, Id-ness etc.)
Secondly step, during the last settings screen, you have to choose between :
-
disable the feature selection
: In this case, all the features retained during the first step will be used to build the model.
-
enable the feature selection
: In this case AutoModel will try to make the model simplier : AutoModel will remove one or several features from the original feature set by using an "automatic feature engineering algorithm" that will test different combinations of features and find the most relevant features among the features of the original features set. As the result, the number of features used to build the model will be less than the number of features in the original features set.
Note that in this last case, AutoModel builds as many models as there are combinations of features tested and thus the computation time is significantly greater than the time with feature selection disabled
I hope it is clearer for you now...
Regards,
Lionel
All comments
lionelderkrikor
Hi
@MikeGer
,
I will answer to the question 3. because your questions 1&2 are incomplete (please complete these questions so that we can understand them...)
The short answer to your question 3. is "not necessarily".
First step, AutoModel select only the relevant feature based on the "quality" of these features (correlation to the target, Id-ness etc.)
Secondly step, during the last settings screen, you have to choose between :
-
disable the feature selection
: In this case, all the features retained during the first step will be used to build the model.
-
enable the feature selection
: In this case AutoModel will try to make the model simplier : AutoModel will remove one or several features from the original feature set by using an "automatic feature engineering algorithm" that will test different combinations of features and find the most relevant features among the features of the original features set. As the result, the number of features used to build the model will be less than the number of features in the original features set.
Note that in this last case, AutoModel builds as many models as there are combinations of features tested and thus the computation time is significantly greater than the time with feature selection disabled
I hope it is clearer for you now...
Regards,
Lionel
MikeGer
Thank you!!! That already helped a lot!!
My computer accidentally posted the same post twice... The complete questions are therefore on another post from my account.
I cant find the option to delete this post, but ADMIN feel free to do so.
Thank you again!
Quick Links
All Categories
Recent Discussions
Activity
Unanswered
日本語 (Japanese)
한국어(Korean)
Groups