Until now I have used LDA only for topic modelling. I would like to know which is the simplest implementation of LDA algorithm for classification tasks.
1 Answer
You can use LDA on your training data to build the topic representation of it for example:
- (document)entry[1] Label A: (topic 1 has 4 words in document 1)T[1]=4, T[3]=7, T[4]=5..
- entry[2] Label C: T[1]=3,T[2]=2...
- entry[3] Label A: T[1]=2,T[2]=2,T[3]=5...
- .
- .
- . -
Using this you can build a simple decision tree: T[1]>1 AND T[3]>4 AND (T[2]>1 OR T[4] > 3) ---> Label A
Another approach would be to use: https://en.wikipedia.org/wiki/Dirichlet-multinomial_distribution#A_second_example:_Naive_Bayes_document_clustering