cross validation - Split dataset to train and test for a LDA model

Question

I have a dataset that contains about 17000 of user data scraped from twitter and I am working with the latent dirichlet allocation algorithm. I want to split my dataset but I am not sure what is the best way. What are the criteria to split a dataset when it comes to train a LDA model. I am using gensim to train LDA model. Thank you

Answer 1

等待大神答复