I'm trying to play around with classification models and started off with logistic regression in R. When I have all the numeric variables in the data set the model works correctly and I was able to interprer the results.

The question is, with another data set I have more than 5 variables which are categorical but they are important though.

1) How do I deal with the variables with categorical values? the total unique categorical values are more than 100 for almost 5 predictors.

2) My dependent variable is having 3 classes, is it okay to using logistic regression for this purpose? (I would still need to explore other classification techniques but right now I'm exploring logistic regression.


1) How often do some of those levels actually appear in your samples? If the outliers (i.e. the infrequent levels) lend little predictive power, you could remove the samples having those ones. Alternatively, you could try binning the categorical levels.

2) Look into multinomial logistic regression.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.