使用igraph包的邻接矩阵

Question

I use igraph package in R for Social Network Analysis. I decide to work with Movielens Dataset (Movies Section) , I also loaded the igraph Library , when I wanted to work with adjacency matrix.

The dataset loaded successfully, and these r my codes.

ff = read.csv("D:/TMU/DataSet/MovieLens/movies.csv", header = TRUE)
ff
mtr = as.matrix(ff)
gr = graph.adjacency(mtr, mode = "undirected", weighted = NULL, diag = FALSE)

I faced with this error :

Error in graph.adjacency.dense(adjmatrix, mode = mode, weighted = weighted, :
At structure_generators.c:274 : Non-square matrix, Non-square matrix
In addition: Warning message:
In mde(x) : NAs introduced by coercion

is there a problem with dataset or what ?

Answer 1

Okay, using the small dataset from https://grouplens.org/datasets/movielens/ which has dimension 9125x3

Download the data (you may need to tweak the mode in the download.file if you are using windows)

pth <- "http://files.grouplens.org/datasets/movielens/ml-latest-small.zip"
download.file(pth, destfile=temp<-tempfile())
#unzip(temp, list=TRUE) # see what files?
unzip(temp, exdir=td<-tempdir()) 

# read movies dataset
movies <- read.csv(file.path(td, "ml-latest-small/movies.csv"), 
                   header=TRUE, stringsAsFactors = FALSE)

Load some libraries

library(tm) # to form the binary matrix: best to keep things sparse
library(slam) # for the crossproduct of the simple_triplet_matrix returned by tm::DocumentTermMatrix
library(igraph) 

Form binary matrix for movies by genres (had to use MrFlick's suggestion of VCorpus otherwise "(no genres listed)" and "film-noir" were split into the individual words

# split the genres string and create binary matrix for presence of genre
corp <- VCorpus(VectorSource(movies$genres))
dtm <- DocumentTermMatrix(corp, 
                          control = list(tokenize = function(x) 
                            unlist(strsplit(as.character(x), "\\|"))))

Create adjacency matrix

# this looks for agreement across the genres
# you could use tcrossprod for similarities on the films
adj <- crossprod_simple_triplet_matrix(dtm)

Create graph

g <- graph_from_adjacency_matrix(adj, mode="undirected", weighted=TRUE)