r - 在 R 中使用余弦距离进行分层聚类

Question

我想通过将余弦相似度与 R 编程语言用于文档语料库来进行层次聚类，但出现以下错误:

Error in if (is.na(n) || n > 65536L) stop("size cannot be NA nor exceed 65536") : missing value where TRUE/FALSE needed

我该怎么办？

为了重现它，这里有一个例子:

library(tm)
doc <- c( "The sky is blue.", "The sun is bright today.", "The sun in the sky is bright.", "We can see the shining sun, the bright sun." )
doc_corpus <- Corpus( VectorSource(doc) )
control_list <- list(removePunctuation = TRUE, stopwords = TRUE, tolower = TRUE)
tdm <- TermDocumentMatrix(doc_corpus, control = control_list)



tf <- as.matrix(tdm)
( idf <- log( ncol(tf) / ( 1 + rowSums(tf != 0) ) ) )
( idf <- diag(idf) )
tf_idf <- crossprod(tf, idf)
colnames(tf_idf) <- rownames(tf)

tf_idf

cosine_dist = 1-crossprod(tf_idf) /(sqrt(colSums(tf_idf^2)%*%t(colSums(tf_idf^2))))
cluster1 <- hclust(cosine_dist, method = "ward.D")

然后我得到错误:

Error in if (is.na(n) || n > 65536L) stop("size cannot be NA nor exceed 65536") : missing value where TRUE/FALSE needed

Answer 1

有2个问题:

1:cosine_dist = 1-crossprod(tf_idf) /(sqrt(colSums(tf_idf^2)%*%t(colSums(tf_idf^2))))创建 NaN 是因为你除以 0。

2:hclust需要一个 dist 对象，而不仅仅是一个矩阵。见 ?hclust更多细节

两者都可以通过以下代码解决:

.....
cosine_dist = 1-crossprod(tf_idf) /(sqrt(colSums(tf_idf^2)%*%t(colSums(tf_idf^2))))

# remove NaN's by 0
cosine_dist[is.na(cosine_dist)] <- 0

# create dist object
cosine_dist <- as.dist(cosine_dist)

cluster1 <- hclust(cosine_dist, method = "ward.D")

plot(cluster1)