I will convert the subtitles into vectors and use them as features to classify the movies into different categories based on their ratings.The problem that I am facing is my feature vector is much much larger compared to the number of examples I have. I would like to know what should the size of my tailing data set be to use LDA,PCA,SVM and naive Bayes.Would 10 movies per category be sufficient?

  • $\begingroup$ As far as I know, SVM should work fine, even if you have small numbers of records per category. But I would try to use more abstract categories. If possible, try to use not more than 10 different categories. $\endgroup$ Dec 25, 2018 at 9:19
  • $\begingroup$ Can you please tell me why you think svm works fine? $\endgroup$ Jan 1, 2019 at 8:51
  • $\begingroup$ "Geometrically, the SVM modeling algorithm works by con- structing a separating hyperplane with the maximal margin. Compared with other standard classifiers, SVM is more accurate on moderately imbalanced data." - Tang, Chawla (2009): SVMs Modeling for Highly Imbalanced Classification. $\endgroup$ Jan 1, 2019 at 16:44


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.