It looks like you're new here. Sign in or register to get started.
I have lot of numerical data that is missing atleast 20+ columns, what type of regression model would work bet to impute those missing values?? I have tried KNN, but results doesnt seem good.??
Also, Under Impute missing values operator when I check "learn on complete cases". I get error stating that the "Example set is empty" . Perhaps I do not have a complete (all values) for any rows of data?? is it really looking at a complete ROW or Column in the dataset? Not sure how to avoid this error. But when i uncheck this the model works and takes ages to output results..
Any insight on this is greatly appreicated.
Yes, complete cases means no missings at all. So if it tells you that exampleset is empty, it means you don't have anyone with all non-missings!
It sounds like you have a lot of missing data, and that can cause serious problems. Do you have some attributes that are more populated than others? You may want to consider narrowing down your modeling attributes to those which are only missing in a few cases.
Without a data sample it is hard to be more specific about which method would work best.