I'm working on a machine learning project aimed at predicting the quality/helpfulness of a review. For each review in the dataset, I have the review text, a number 'm' for the number of people who have voted on the review and a number 'n' for the number of positive votes on the review.

The goal is to predict the percentage of votes that are positive:


I'm using a random forest for the main algorithm, and trying to decide on what would be a good algorithm to use for the baseline.

A feature vector for each review comprises a word presence representation of the review and a number representing the total number of words in the review.

I would appreciate any suggestions on what algorithm would be good for a baseline method to compare against my random forest implementation.


  • $\begingroup$ A (simple) logistic regression would be the obvious choice for a binary outcome. It should be relatively easy to implement too. $\endgroup$ – usεr11852 Mar 8 '15 at 7:46
  • $\begingroup$ @usεr11852 - it's not a binary outcome though. Say 35 people have voted on the review, out of which 20 thought the review was helpful. The label for this particular example is (20/35 = 0.5714) $\endgroup$ – Stralo Mar 8 '15 at 7:51
  • $\begingroup$ Yeap! You won't get back a binary prediction from the logistic regression. It returns a number between 0 and 1. Check the logit() and probit() link functions. $\endgroup$ – usεr11852 Mar 8 '15 at 8:35

Your Answer

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

Browse other questions tagged or ask your own question.