There is a dataset with about 20 dimensional input vectors and binary output vector. There are about 100K training examples. When trained using logistic regression the accuracy on training set is about 50% and trained using random forest classifier(100 trees) the accuracy on training set is 100%. It is a general notion that random forest classifiers are less prone to overfitting. But in this particular case, the test accuracy of the logistic regression classifier is much higher than the random forest classifier. What can we infer about the distribution of the dataset from this observation?

  • $\begingroup$ What's the test error of both algorithms? $\endgroup$ – horaceT Jun 1 '16 at 19:02
  • $\begingroup$ We need more details like the test error, the covariates selected by the random forest, the percentage of trees behind the 100% success rate, &tc. Have you checked the results when switching testing and training sets? $\endgroup$ – Xi'an Jun 1 '16 at 19:46
  • $\begingroup$ The test accuracy is ~50% for logistic regression and ~30% for Random Forest. I need to do some more work to find out the percentage of trees behind the 100% success rate. Will get back on this. $\endgroup$ – web_ninja Jun 1 '16 at 20:06

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.