Mastering Predictive Analytics with R

Latent Dirichlet Allocation

Latent Dirichlet Allocation (LDA) is the prototypical method to perform topic modeling. Rather unfortunately, the acronym LDA is also used for another method in machine learning, Linear Discriminant Analysis. This latter method is completely different to Latent Dirichlet Allocation and is commonly used as a way to perform dimensionality reduction and classification. Needless to say, we will use LDA to refer to Latent Dirichlet Allocation throughout this book.

Although LDA involves a substantial amount of mathematics, it is worth exploring some of its technical details in order to understand how the model works and the assumptions that it uses. First and foremost, we should learn about the Dirichlet distribution, which ...