Data mining techniques - Selecting important attributes

In reality there are many attributes in data, but all are not important. So, we have to find the important attributes for analysis.

So you might be thinking now that are there any methods for attribute selection in data mining?

Yes, there are several techniques, depending on the types of modeling you are doing.

In general, you may follow following steps as mentioned in order.

1) Reduce variables as per missing percentage.

2) Reduce variables having high VIF (multicollinerity check)

3) If some Xs are highly correlated with Ys, drop them at the beginning of the modeling.

4) Reduce variables as per information value.

5) Reduce variables as per significance Retain P value<0.0001.

6) Out of significant variables, reduce variables having lower |t| or low Chi sq.

7) Drop variables whose signs are "lurking", i.e., sign not same when they are individually introduced in model compared to when they are put together into a model. In both scenarios, sign/magnitude of parameter estimates should be same though estimated value will be different.

8) If still you have "more" variables, then you may try to use factor analysis to reduce number of variables and instead of those, use liner combination of those variables, i.e., by using factor.

9) Last but not the least, reduce variables, which don't make business sense.

Data mining is important for your business and you should know the attibutes properly

Data mining helps bring more unique visitors
Data mining helps bring more unique visitors
Visual data mining
Visual data mining
Data mining and domestic security
Data mining and domestic security
data mining techniques
data mining techniques
Web data mines
Web data mines

Data Mining - a short video introduction to professors

More by this Author


Comments 1 comment

4 years ago

good

    Sign in or sign up and post using a HubPages Network account.

    0 of 8192 characters used
    Post Comment

    No HTML is allowed in comments, but URLs will be hyperlinked. Comments are not for promoting your articles or other sites.


    Click to Rate This Article
    working