R/z_runAlgorithm.R

   1 #' @include b_Algorithm.R
   2
   3 algoNameDictionary = list(
   4     ew = "ExponentialWeights",
   5     kn = "KnearestNeighbors",
   6     ga = "GeneralizedAdditive",
   7     ml = "MLpoly",
   8     rt = "RegressionTree",
   9     rr = "RidgeRegression",
  10     sv = "SVMclassif"
  11 )
  12
  13 #' @title Simulate real-time predict
  14 #'
  15 #' @description Run the algorithm coded by \code{shortAlgoName} on data specified by the \code{stations} argument.
  16 #'
  17 #' @param shortAlgoName Short name of the algorithm.
  18 #' \itemize{
  19 #'   \item ew : Exponential Weights
  20 #'   \item ga : Generalized Additive Model
  21 #'   \item kn : K Nearest Neighbors
  22 #'   \item ml : MLpoly
  23 #'   \item rt : Regression Tree
  24 #'   \item rr : Ridge Regression
  25 #' }
  26 #' @param experts Vector of experts to consider (names or indices). Default: all of them.
  27 #' @param stations Vector of stations to consider (names or indices). Default: all of them.
  28 #' @param ... Additional arguments to be passed to the Algorithm object.
  29 #'
  30 #' @return A list with the following slots
  31 #' \itemize{
  32 #'   \item{data : data frame of all forecasts + measures (may contain NAs) + predictions, with date and station indices.}
  33 #'   \item{algo : object of class \code{Algorithm} (or sub-class).}
  34 #'   \item{experts : character vector of experts for this run.}
  35 #'   \item{stations : character vector of stations for this run.}
  36 #' }
  37 #'
  38 #' @export
  39 runAlgorithm = function(shortAlgoName, experts=expertsArray, stations=stationsArray, ...)
  40 {
  41     #check, sanitize and format provided arguments
  42     if (! shortAlgoName %in% names(algoNameDictionary))
  43         stop(paste("Typo in short algo name:", shortAlgoName))
  44     if (!is.character(experts) && !is.numeric(experts))
  45         stop("Wrong argument type: experts should be character or integer")
  46     if (!is.character(stations) && !is.numeric(stations))
  47         stop("Wrong argument type: stations should be character or integer")
  48     experts = unique(experts)
  49     stations = unique(stations)
  50     Ka = length(expertsArray)
  51     Sa = length(stationsArray)
  52     if (length(experts) > Ka)
  53         stop("Too many experts specified: at least one of them does not exist")
  54     if (length(stations) > Sa)
  55         stop("Too many stations specified: at least one of them does not exist")
  56     if (is.numeric(experts) && any(experts > Ka))
  57         stop(paste("Some experts indices are higher than the maximum which is", Ka))
  58     if (is.numeric(stations) && any(stations > Sa))
  59         stop(paste("Some stations indices are higher than the maximum which is", Sa))
  60     if (is.character(experts))
  61     {
  62         expertsMismatch = (1:Ka)[! experts %in% expertsArray]
  63         if (length(expertsMismatch) > 0)
  64             stop(cat(paste("Typo in experts names:", experts[expertsMismatch]), sep="\n"))
  65     }
  66     if (is.character(stations))
  67     {
  68         stationsMismatch = (1:Sa)[! stations %in% stationsArray]
  69         if (length(stationsMismatch) > 0)
  70             stop(cat(paste("Typo in stations names:", stations[stationsMismatch]), sep="\n"))
  71     }
  72     if (!is.character(experts))
  73         experts = expertsArray[experts]
  74     if (!is.character(stations))
  75         stations = stationsArray[stations]
  76
  77     #get data == ordered date indices + forecasts + measures + stations indices (would be DB in prod)
  78     oracleData = getData(experts, stations)
  79
  80     #simulate incremental forecasts acquisition + prediction + get measure
  81     algoData = as.data.frame(matrix(nrow=0, ncol=ncol(oracleData)))
  82     names(algoData) = names(oracleData)
  83     algorithm = new(algoNameDictionary[[shortAlgoName]], data=algoData, ...)
  84     predictions = c()
  85     T = oracleData[nrow(oracleData),"Date"]
  86     for (t in 1:T)
  87     {
  88         #NOTE: bet that subset extract rows in the order they appear
  89         tData = subset(oracleData, subset = (Date==t))
  90         algorithm$inputNextForecasts(tData[,names(tData) != "Measure"])
  91         predictions = c(predictions, algorithm$predict_withNA())
  92         algorithm$inputNextObservations(tData[,"Measure"])
  93     }
  94
  95     oracleData = cbind(oracleData, Prediction = predictions)
  96     return (list(data = oracleData, algo = algorithm, experts = experts, stations = stations))
  97 }