old_C_code/stage2_UNFINISHED/src/unused/05_cluster2stepWER-par.r

   1 ## File : 05_cluster2stepWER.r
   2 ## Description :
   3
   4 rm(list = ls())
   5
   6 if(Sys.info()[4] ==  "mojarrita"){
   7   setwd("~/Documents/projects/2014_EDF-Orsay-Lyon2/codes/")
   8 } else {
   9   setwd("~/ownCloud/projects/2014_EDF-Orsay-Lyon2/codes/")
  10 }
  11
  12 library(Rwave)       # CWT
  13 library(cluster)     # pam
  14 #library(flexclust)   # kcca
  15 source("aux.r")               # auxiliary clustering functions
  16 source("sowas-superseded.r")  # auxiliary CWT functions
  17
  18 ## 1. Read auxiliar data files ####
  19
  20 identifiants <- read.table("identifs.txt")[ ,1]
  21 dates0       <- read.table("datesall.txt")[, 1]
  22 dates        <- as.character(dates0[grep("2009", dates0)])
  23 rm(dates0)
  24
  25 #n <- length(identifiants)
  26 #p <- length(dates)
  27
  28 load('~/tmp/2009synchrosdf200WER')
  29 #load('../res/2009_synchros200WER.Rdata')
  30 synchros09 <- synchros
  31 load('~/tmp/2010synchrosdf200WER')
  32 synchros10 <- synchros; rm(synchros)
  33
  34 conso <- lapply(synchros09, function(ll) {
  35                 nas <- which(is.na(ll)[, 1]) # some 1/1/2009 are missing
  36                 ll[nas, 1] <- rowMeans(ll[nas, 2:4])
  37                 imput <- ll[, 4180:4181] %*% matrix(c(2/3, 1/3, 1/3, 2/3), 2)
  38                 cbind(ll[, 1:4180], imput, ll[, 4181:17518]) } )
  39
  40 n     <- nrow(conso[[1]])
  41 delta <- ncol(conso[[1]])
  42 rm(synchros09)
  43
  44
  45 ## 2. Compute WER distance matrix ####
  46
  47 ## _.a CWT -- Filtering the lowest freqs (>6m) ####
  48 nvoice   <- 4
  49 # noctave4 = 2^13 = 8192 half hours ~ 180 days
  50 noctave4 <- adjust.noctave(N = delta, dt = 1, s0 = 2,
  51                            tw = 0, noctave = 13)
  52 # 4 here represent 2^5 = 32 half-hours ~ 1 day
  53 scalevector4  <- 2^(4:(noctave4 * nvoice) / nvoice) * 2
  54 lscvect4      <- length(scalevector4)
  55 lscvect <- lscvect4  # i should clean my code: werFam demands a lscvect
  56
  57 Xcwt4 <- lapply(conso, function(ll)
  58   toCWT(ll, noctave = noctave4, dt = 1,
  59         scalevector = scalevector4,
  60         lt = delta, smooth = FALSE,
  61         nvoice = nvoice)      # observations node with CWT
  62   )
  63
  64 rm(conso)
  65
  66 Xcwt4 <- lapply(conso, function(ll) {
  67   aux <- toCWT(ll, noctave = noctave4, dt = 1,
  68                scalevector = scalevector4,
  69                lt = delta, smooth = FALSE,
  70                nvoice = nvoice)      # observations node with CWT
  71   res <- matrix(NA_complex_, nrow = n, ncol= 2 + length((c(aux[,,1]))))
  72   for(i in 1:n)
  73     res[i, ] <- c(delta, lscvect, res[,,i] / max(Mod(res[,,i])) )
  74   res })
  75
  76 rm(conso)
  77
  78
  79 #Xcwt2 <- matrix(0.0, nrow= n, ncol= 2 + delta * lscvect)
  80 #Xcwt2 <- matrix(NA_complex_, nrow= n, ncol= 2 + length((c(Xcwt4[,,1]))))
  81
  82 #for(i in 1:n)
  83 #  Xcwt2[i,] <- c(delta, lscvect, Xcwt4[,,i] / max(Mod(Xcwt4[,,i])) )
  84
  85 #rm(conso, Xcwt4); gc()
  86
  87 ## _.b WER^2 distances  ########
  88 #Xwer_dist    <- matrix(0.0, n, n)
  89 #for(i in 1:(n - 1)){
  90 #  mat1   <- vect2mat(Xcwt2[i,])
  91 #  for(j in (i + 1):n){
  92 #     mat2 <- vect2mat(Xcwt2[j,])
  93 #     num     <- Mod(mat1 * Conj(mat2))
  94 #     WX      <- Mod(mat1 * Conj(mat1))
  95 #     WY      <- Mod(mat2 * Conj(mat2))
  96 #     smsmnum <- smCWT(num, scalevector = scalevector4)
  97 #     smsmWX  <- smCWT(WX,  scalevector = scalevector4)
  98 #     smsmWY  <- smCWT(WY,  scalevector = scalevector4)
  99 #     wer2    <- sum(colSums(smsmnum)^2)  /
 100 #       sum( sum(colSums(smsmWX) * colSums(smsmWY)) )
 101 #     Xwer_dist[i, j] <- sqrt(delta * lscvect * (1 - wer2))
 102 #     Xwer_dist[j, i] <- Xwer_dist[i, j]
 103 #   }
 104 # }
 105 # diag(Xwer_dist) <- numeric(n)
 106 #
 107 # save(Xwer_dist, file = "../res/2009_synchros200WER.Rdata")
 108
 109 load("../res/2009_synchros200WER.Rdata")
 110
 111
 112 ## 3. Cluster using WER distance matrix ####
 113
 114 #hc    <- hclust(as.dist(Xwer_dist), method = "ward.D")
 115 #plot(hc)
 116 #
 117 # #clust <- cutree(hc, 2)
 118 #
 119 for(K in 2:30){
 120   #K <- 3
 121   #pamfit <- pam(tdata[-201, ci$selectv], k = K)
 122   pamfit <- pam(as.dist(Xwer_dist), k = K, diss = TRUE)
 123
 124   #table(pamfit$clustering)
 125
 126   SC <- matrix(0, ncol = p, nrow = K)
 127
 128   clustfactor <- pamfit$clustering
 129 #  for(k in 1:K){
 130 #    clustk <- which(clustfactor == k)
 131 #  if(length(clustk) > 0) {
 132 #    if(length(clustk) > 1) {
 133 #      SCk <- colSums(synchros09[which(clustfactor == k), ])
 134 #      } else {
 135 #        SCk <- synchros09[which(clustfactor == k), ]
 136 #      }
 137 #      SC[k, ] <- SC[k, ] + SCk
 138 #      rm(SCk)
 139 #  }
 140 #}
 141
 142 #write.table(clustfactor, file = paste0("~/tmp/clustfactorRC", K, ".txt"))
 143 #write.table(clustfactor, file = "~/tmp/clustfactor3.txt")
 144 write.table(clustfactor, file = paste0("~/tmp/clustfactorWER", K, ".txt"))
 145 }
 146 #
 147 # # Plots
 148 # layout(1)
 149 # matplot(t(SC)[48*10 + 1:(48*30), ],  type = 'l', ylab = '',col = 1:3, lty = 1)
 150 # matplot(t(SC)[48*100 + 1:(48*30), ], type = 'l', ylab = '', col = 1:3, lty = 1)
 151 #
 152 #
 153 #