ldaRobust: LDA Robustness

#' rlda_c
#'
#' Main class
#'
#' methods: LDA wrapper, fit
#'
#'
#' @import topicmodels
#' @import foreach
#' @import doParallel
#' @import parallel
#' @import ggplot2
#' @import reshape2
#' @import dplyr
#' @import tidyr
#' @import lsa
#' @importClassesFrom topicmodels LDA
#' @import tm
#' @import Hmisc
#' @import weights
#' @slot  dtm document term matrix
#' @slot lda_u lda object(in topicmodels)
#' @slot K numeric or vector, if numeric, number of k to try, if vector, k's to try (will be overwrite with list of k's tried once fit has been run)
#' @slot compute_parallel bool, if true, will parallel when fitting lda models; otherwise, sequential for fitting
#' @slot same_k_estimation integer, number of initial states to try, if 0, do not try with diff state
#' @slot similarity_measure string, similarity measure (so far cosine or hellinger). Default: cosine
#' @slot num_of_clusters numeric or vector, number of clusters used when performing spectral clustering
#' @slot seed_list seeds tried (exists only when same_k_estimation is true)
#' @slot beta_list list of beta matrices in LDA objects of all K tried (ordered same as K)
#' @slot gamma_list list of gamma matrices in LDA objects of all K tried (ordered same as K)
#' @slot model_topic_mat percentage of documents dominated by the given topic
#' @slot similarity_mat maximum similarity (given choice of similarity functions) of a given topic compare to any topics in the original lda model (to give the probability of a user’s topic shows up in a tried model’s resulting topics)
#' @slot sim_matrix_list list of similarity matrices that gives us similarity between
#' @slot init_states # not used yet, probably gonna delete
#' @slot key_features top 10 features of a given topic in each model tried
#' @slot topic_dom_perc_list percentage of documents dominated by the given topic out of documents originally dominated by similar topic in the original model
#' @slot dominant_topic_cluster_list clusters correponding to dominant topics of each document in each model
#' @slot cluster_center_key_words_list (data_frame?) top 10 keywords for each center found by the cluster algorithn (so far only support spectral clustering)
#' @slot perc_document_belong_cluster_list percentage of documents belong to a given cluster in a given model
#' @slot topic_cluster_assignment cluster number a given topic belongs to d
#' @slot top_stability_mat a matrix indicating whether each cluster is present in each alternative model
#' @slot doc_by_cluster_and_model a matrix indicating the dominiant cluster of each document according to each topic model
#' @exportClass rlda


#setClassUnion("Nul_meric", c("numeric", "NULL"))
#setClassUnion("Nul_DA", c("LDA", "NULL"))
#devtools::use_package("weights")
#devtools::use_package("Hmisc")


setClass("rlda",
         representation(dtm="DocumentTermMatrix",
                        lda_u = "LDA",
                        K = "numeric",
                        model_type = "character",
                        other_dtms = "list",
                        threshold="numeric",
                        same_k_estimation="numeric",
                        compute_parallel = "logical",
                        similarity_measure="character",
                        num_of_clusters = "numeric",
                        seed_list = "numeric", #?????,
                        #model_list = "list",
                        beta_list = "list",
                        gamma_list = "list",
                        model_topic_mat = "list",
                        similarity_mat = "list",
                        similarity_mat_list = "list",
                        init_states = "numeric",
                        topic_dom_perc_list = "list",
                        dominant_topic_cluster_list = "list",
                        key_features = "list",
                        terms = "character",
                        cluster_center_key_words_list = "data.frame",
                        perc_document_belong_cluster_list = "list",
                        topic_cluster_assignment = "matrix",
                        top_stability_mat = "data.frame",
                        docs_by_cluster_and_model = "data.frame"
         ),
         prototype(K=5,
                   #threshold=0.5,
                   same_k_estimation=0,
                   similarity_measure="cosine",
                   compute_parallel = TRUE,
                   #seed_list = NULL,
                   #model_list = NULL,
                   #model_topic_mat =NULL,
                   #similarity_mat = NULL,
                   #similarity_mat_list = NULL,
                   init_states = 0)
)