I hope we can remove the highly correlated variables based on the feature importance may be with PCA etc.
Is there anything we can do with highly correlated variables/
Thanks in advance !
Welcome to Data Science Stack Exchange
Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It's built and run by you as part of the Stack Exchange network of Q&A sites. With your help, we're working together to build a library of detailed answers to every question about data science.
We're a little bit different from other sites. Here's how:
Ask questions, get answers, no distractions
This site is all about getting answers. It's not a discussion forum. There's no chit-chat.
Just questions...
...and answers.
up vote
Good answers are voted up and rise to the top.
The best answers show up first so that they are always easy to find.
accept
The person who asked can mark one answer as "accepted".
Accepting doesn't mean it's the best answer, it just means that it worked for the person who asked.
What can be done with highly correlated variables (>.95 and <-.95)
up vote
14
down vote
asked Feb 7 '20 at 12:56
Vivek Ananthan
2052 silver badges8 bronze badges
2052 silver badges8 bronze badges
2 Answers
up vote
4
down vote
accept
You need to remove them. Redundant features only increase the computation time, increase model complexity (with no benefit) which means making interpretation of model/analysis more sophisticated and if they are many, removing them prunes your vector space by improving the density of information in dimensions of vector space (it helps e.g. in finding nearest neighbors).
answered Feb 7 '20 at 16:08
Kasra Manshaei
5,26113 silver badges38 bronze badges
5,26113 silver badges38 bronze badges
up vote
3
down vote
I think merging such correlated features and create a new one, will also be a good idea. In that way we will not lose any information.
For example, sum up the values of different correlated features and take an average of it, will be the very basic option.
answered Feb 7 '20 at 17:44
vipin bansal
1,2622 silver badges13 bronze badges
1,2622 silver badges13 bronze badges
Get answers to practical, detailed questions
Focus on questions about an actual problem you have faced. Include details about what you have tried and exactly what you are trying to do.
Ask about...
- Specific issues with data science
- Real problems or questions that you’ve encountered
Not all questions work well in our format. Avoid questions that are primarily opinion-based, or that are likely to generate discussion rather than answers.
Questions that need improvement may be closed until someone fixes them.
Don't ask about...
- Anything not directly related to data science
- Questions that are primarily opinion-based
- Questions with too many possible answers or that would require an extremely long answer
Tags make it easy to find interesting questions
All questions are tagged with their subject areas. Each can have up to 5 tags, since a question might be related to several subjects.
Click any tag to see a list of questions with that tag, or go to the tag list to browse for topics that interest you.
What can be done with highly correlated variables (>.95 and <-.95)
up vote
14
down vote
I hope we can remove the highly correlated variables based on the feature importance may be with PCA etc.
Is there anything we can do with highly correlated variables/
Thanks in advance !
asked Feb 7 '20 at 12:56
Vivek Ananthan
2052 silver badges8 bronze badges
2052 silver badges8 bronze badges
You earn reputation when people vote on your posts
Your reputation score goes up when others vote up your questions, answers and edits.
+10
question voted up
+10
answer voted up
+15
answer is accepted
+2
edit approved
As you earn reputation, you'll unlock new privileges like the ability to vote, comment, and even edit other people's posts.
| Reputation | Privilege |
|---|---|
| 15 | Vote up |
| 50 | Leave comments |
| 125 | Vote down (costs 1 rep on answers) |
At the highest levels, you'll have access to special moderation tools. You'll be able to work alongside our community moderators to keep the site focused and helpful.
| 500 | Vote to close, reopen, or migrate questions |
|---|---|
| 1000 | Edit other people's posts |
| 2000 | Access to moderation tools |
Improve posts by editing or commenting
Our goal is to have the best answers to every question, so if you see questions or answers that can be improved, you can edit them.
Use edits to fix mistakes, improve formatting, or clarify the meaning of a post.
up vote
9
down vote
You need to remove them. Redundant features only increase the computation time, increase model complexity (with no benefit) which means making interpretation of model/analysis more sophisticated and if they are many, removing them prunes your vector space by improving the density of information in dimensions of vector space (it helps e.g. in finding nearest neighbors).
answered Feb 7 '20 at 16:08
Kasra Manshaei
5,26113 silver badges38 bronze badges
5,26113 silver badges38 bronze badges
In the situation of two highly correlated features, actually choosing one does the same job. Question mentioned .95 which means they are practically the same. Plus the fact that merging strategy needs to be chosen. What is ur idea for merging? - Kasra Manshaei Feb 7 '20 at 17:44
Unlock badges for special achievements
Badges are special achievements you earn for participating on the site. They come in three levels: bronze, silver, and gold.
| Student | First question with score of 1 or more |
| Editor | First edit |
| Good Answer | Answer score of 25 or more |
| Civic Duty | Vote 300 or more times |
| Famous Question | Question with 10,000 views |
Sign up to get started
Signing up allows you to:
- Earn reputation when you help others with questions, answers and edits.
- Select favorite tags to customize your home page.
- Claim your first badge: Informed
Looking for more in-depth information on the site? Visit the Help Center
Data Science Stack Exchange is part of the Stack Exchange network
Like this site? Stack Exchange is a network of 176 Q&A sites just like it. Check out the full list of sites.
Use comments to ask for more information or clarify a question or answer.
You can always comment on your own questions and answers. Once you earn 50 reputation, you can comment on anybody's post.
Remember: we're all here to learn, so be friendly and helpful!