fit.topics.perplexity: fit.topics.perplexity

In bvidgen/RPackage: Big Data Analytics functions

Description

Calculates perplexity for values of k on an LDA topic model using the topicmodels package, split into training and testing sets using k-folds

Usage

fit.topics.perplexity(dtm, folds, k.values, alpha, beta, control.test)

Arguments

dtm	Document-term matrix. Constructed using the DocumentTermMatrix() command from the tm package
folds	Integer. The number of folds to make training and testing sets; recommended values are '5' and '10' - note that higher values considerably increase the time that model fitting takes
k.values	Numeric vector. Values to test k for. A good starting point is 2:10. All values must be greater than 1
alpha	Numeric. Optional parameter: the value of alpha used in the LDA model. By default alpha is set to 0.1
beta	Numeric. Optional parameter: the value of beta used in the LDA model. By default beta is set to 0.1
control.test	List. Optional parameter: the LDA control list used in the LDA model. It is strongly recommended not to use this parameter unless you have good reason. Default settings are: nstart = 5, best = T, burnin = 1000, iter = 2000, thin = 500

Value

Dataframe of perplexity for the k.values, calculated for the number of stipulated folds

bvidgen/RPackage documentation built on May 13, 2019, 9:04 a.m.