聚類分析有不少種, 效果好很差大概要根據數據特徵來肯定。最多見的是kmeans法聚類html
> setwd("D:\\R_test") > data_in <- read.delim("tmp_result.txt", header=T) > fit <- kmeans(data_in, 3) > library(cluster) > clusplot(data_in, fit$cluster, color=T, shade=T, labels = 2, lines =0)
也能夠用mclustwindows
> install.packages("mclust") 試開URL’http://cloud.r-project.org/bin/windows/contrib/2.14/mclust_4.0.zip' Content type 'application/zip' length 2371233 bytes (2.3 Mb) 打開了URL downloaded 2.3 Mb 程序包‘mclust’打開成功,MD5和檢查也經過 下載的程序包在 C:\Users\Administrator\AppData\Local\Temp\RtmpiIyw2o\downloaded_packages裏 > fit <- Mclust(data_in) > summary(fit) ---------------------------------------------------- Gaussian finite mixture model fitted by EM algorithm ---------------------------------------------------- Mclust XXX (elliposidal multivariate normal) model with 1 component: log.likelihood n df BIC 1616504 263 33410 3046843 Clustering table: 1 263 > fit$ // 按下Tab鍵,有如下選項 fit$call fit$modelName fit$n fit$d fit$G fit$BIC fit$bic fit$loglik fit$df fit$parameters fit$classification fit$uncertainty > plot(fit, what="classification") // http://www.statmethods.net/advstats/cluster.html