epclust/R/main.R

   1 #' @include defaults.R
   2
   3 #' @title Cluster power curves with PAM in parallel
   4 #'
   5 #' @description Groups electricity power curves (or any series of similar nature) by applying PAM
   6 #' algorithm in parallel to chunks of size \code{nb_series_per_chunk}
   7 #'
   8 #' @param data Access to the data, which can be of one of the three following types:
   9 #' \itemize{
  10 #'   \item data.frame: each line contains its ID in the first cell, and all values after
  11 #'   \item connection: any R connection object (e.g. a file) providing lines as described above
  12 #'   \item function: a custom way to retrieve the curves; it has two arguments: the start index
  13 #'     (start) and number of curves (n); see example in package vignette.
  14 #' }
  15 #' @param K Number of clusters
  16 #' @param nb_series_per_chunk (Maximum) number of series in each group
  17 #' @param min_series_per_chunk Minimum number of series in each group
  18 #' @param writeTmp Function to write temporary wavelets coefficients (+ identifiers);
  19 #'   see defaults in defaults.R
  20 #' @param readTmp Function to read temporary wavelets coefficients (see defaults.R)
  21 #' @param wf Wavelet transform filter; see ?wt.filter. Default: haar
  22 #' @param WER "end" to apply stage 2 after stage 1 has iterated and finished, or "mix"
  23 #'   to apply it after every stage 1
  24 #' @param ncores number of parallel processes; if NULL, use parallel::detectCores()
  25 #'
  26 #' @return A data.frame of the final medoids curves (identifiers + values)
  27 epclust = function(data, K, nb_series_per_chunk, min_series_per_chunk=10*K,
  28     writeTmp=defaultWriteTmp, readTmp=defaultReadTmp, wf="haar", WER="end", ncores=NULL)
  29 {
  30     #TODO: setRefClass(...) to avoid copy data:
  31     #http://stackoverflow.com/questions/2603184/r-pass-by-reference
  32
  33     #0) check arguments
  34     if (!is.data.frame(data) && !is.function(data))
  35         tryCatch(
  36             {
  37                 if (is.character(data))
  38                 {
  39                     data_con = file(data, open="r")
  40                 } else if (!isOpen(data))
  41                 {
  42                     open(data)
  43                     data_con = data
  44                 }
  45             },
  46             error="data should be a data.frame, a function or a valid connection")
  47     if (!is.integer(K) || K < 2)
  48         stop("K should be an integer greater or equal to 2")
  49     if (!is.integer(nb_series_per_chunk) || nb_series_per_chunk < K)
  50         stop("nb_series_per_chunk should be an integer greater or equal to K")
  51     if (!is.function(writeTmp) || !is.function(readTmp))
  52         stop("read/writeTmp should be functional (see defaults.R)")
  53     if (WER!="end" && WER!="mix")
  54         stop("WER takes values in {'end','mix'}")
  55     #concerning ncores, any non-integer type will be treated as "use parallel:detectCores()"
  56
  57     #1) acquire data (process curves, get as coeffs)
  58     #TODO: for data.frame and custom function, run in parallel (connections are sequential[?!])
  59     index = 1
  60     nb_curves = 0
  61     repeat
  62     {
  63         coeffs_chunk = NULL
  64         if (is.data.frame(data))
  65         {
  66             #full data matrix
  67             if (index < nrow(data))
  68             {
  69                 coeffs_chunk = curvesToCoeffs(
  70                     data[index:(min(index+nb_series_per_chunk-1,nrow(data))),], wf)
  71             }
  72         } else if (is.function(data))
  73         {
  74             #custom user function to retrieve next n curves, probably to read from DB
  75             coeffs_chunk = curvesToCoeffs( data(index, nb_series_per_chunk), wf )
  76         } else
  77         {
  78             #incremental connection
  79             #TODO: find a better way to parse than using a temp file
  80             ascii_lines = readLines(data_con, nb_series_per_chunk)
  81             if (length(ascii_lines > 0))
  82             {
  83                 series_chunk_file = ".tmp/series_chunk"
  84                 writeLines(ascii_lines, series_chunk_file)
  85                 coeffs_chunk = curvesToCoeffs( read.csv(series_chunk_file), wf )
  86             }
  87         }
  88         if (is.null(coeffs_chunk))
  89             break
  90         writeTmp(coeffs_chunk)
  91         nb_curves = nb_curves + nrow(coeffs_chunk)
  92         index = index + nb_series_per_chunk
  93     }
  94     if (exists(data_con))
  95         close(data_con)
  96     if (nb_curves < min_series_per_chunk)
  97         stop("Not enough data: less rows than min_series_per_chunk!")
  98
  99     #2) process coeffs (by nb_series_per_chunk) and cluster them in parallel
 100     library(parallel)
 101     ncores = ifelse(is.integer(ncores), ncores, parallel::detectCores())
 102     cl = parallel::makeCluster(ncores)
 103     parallel::clusterExport(cl=cl, varlist=c("TODO:", "what", "to", "export?"), envir=environment())
 104     #TODO: be careful of writing to a new temp file, then flush initial one, then re-use it...
 105     repeat
 106     {
 107         #while there is jobs to do (i.e. size of tmp "file" is greater than nb_series_per_chunk)
 108         nb_workers = nb_curves %/% nb_series_per_chunk
 109         indices = list()
 110         #indices[[i]] == (start_index,number_of_elements)
 111         for (i in 1:nb_workers)
 112             indices[[i]] = c(nb_series_per_chunk*(i-1)+1, nb_series_per_chunk)
 113         remainder = nb_curves %% nb_series_per_chunk
 114         if (remainder >= min_series_per_chunk)
 115         {
 116             nb_workers = nb_workers + 1
 117             indices[[nb_workers]] = c(nb_curves-remainder+1, nb_curves)
 118         } else if (remainder > 0)
 119         {
 120             #spread the load among other workers
 121             #...
 122         }
 123         li = parallel::parLapply(cl, indices, processChunk, K, WER=="mix")
 124         #C) flush tmp file (current parallel processes will write in it)
 125     }
 126     parallel::stopCluster(cl)
 127
 128     #3) readTmp last results, apply PAM on it, and return medoids + identifiers
 129     final_coeffs = readTmp(1, nb_series_per_chunk)
 130     if (nrow(final_coeffs) == K)
 131     {
 132         return ( list( medoids=coeffsToCurves(final_coeffs[,2:ncol(final_coeffs)]),
 133             ids=final_coeffs[,1] ) )
 134     }
 135     pam_output = getClusters(as.matrix(final_coeffs[,2:ncol(final_coeffs)]), K)
 136     medoids = coeffsToCurves(pam_output$medoids, wf)
 137     ids = final_coeffs[,1] [pam_output$ranks]
 138
 139     #4) apply stage 2 (in parallel ? inside task 2) ?)
 140     if (WER == "end")
 141     {
 142         #from center curves, apply stage 2...
 143         #TODO:
 144     }
 145
 146     return (list(medoids=medoids, ids=ids))
 147 }
 148
 149 processChunk = function(indice, K, WER)
 150 {
 151     #1) retrieve data
 152     coeffs = readTmp(indice[1], indice[2])
 153     #2) cluster
 154     cl = getClusters(as.matrix(coeffs[,2:ncol(coeffs)]), K)
 155     #3) WER (optional)
 156     #TODO:
 157 }
 158
 159 #TODO: difficulté : retrouver courbe à partir de l'identifiant (DB ok mais le reste ?)
 160 #aussi : que passe-t-on aux noeuds ? curvesToCoeffs en // ?
 161 #enfin : WER ?!