pkg/R/F_Neighbors.R

   1 #' @include Forecaster.R
   2 #'
   3 #' Neighbors Forecaster
   4 #'
   5 #' Predict tomorrow as a weighted combination of "futures of the past" days.
   6 #' Inherits \code{\link{Forecaster}}
   7 #'
   8 NeighborsForecaster = R6::R6Class("NeighborsForecaster",
   9     inherit = Forecaster,
  10
  11     public = list(
  12         predictShape = function(data, today, memory, horizon, ...)
  13         {
  14             # (re)initialize computed parameters
  15             private$.params <- list("weights"=NA, "indices"=NA, "window"=NA)
  16
  17             # Do not forecast on days with NAs (TODO: softer condition...)
  18             if (any(is.na(data$getCenteredSerie(today))))
  19                 return (NA)
  20
  21             # Determine indices of no-NAs days followed by no-NAs tomorrows
  22             fdays = getNoNA2(data, max(today-memory,1), today-1)
  23
  24             # Get optional args
  25             local = ifelse(hasArg("local"), list(...)$local, FALSE) #same level + season?
  26             simtype = ifelse(hasArg("simtype"), list(...)$simtype, "mix") #or "endo", or "exo"
  27             if (hasArg("window"))
  28             {
  29                 return ( private$.predictShapeAux(data,
  30                     fdays, today, horizon, local, list(...)$window, simtype, TRUE) )
  31             }
  32
  33             # Indices of similar days for cross-validation; TODO: 20 = magic number
  34             cv_days = getSimilarDaysIndices(today, data, limit=20, same_season=FALSE,
  35                 days_in=fdays)
  36
  37             # Optimize h : h |--> sum of prediction errors on last 45 "similar" days
  38             errorOnLastNdays = function(window, simtype)
  39             {
  40                 error = 0
  41                 nb_jours = 0
  42                 for (i in seq_along(cv_days))
  43                 {
  44                     # mix_strategy is never used here (simtype != "mix"), therefore left blank
  45                     prediction = private$.predictShapeAux(data, fdays, cv_days[i], horizon, local,
  46                         window, simtype, FALSE)
  47                     if (!is.na(prediction[1]))
  48                     {
  49                         nb_jours = nb_jours + 1
  50                         error = error +
  51                             mean((data$getSerie(cv_days[i]+1)[1:horizon] - prediction)^2)
  52                     }
  53                 }
  54                 return (error / nb_jours)
  55             }
  56
  57             if (simtype != "endo")
  58             {
  59                 best_window_exo = optimize(
  60                     errorOnLastNdays, c(0,7), simtype="exo")$minimum
  61             }
  62             if (simtype != "exo")
  63             {
  64                 best_window_endo = optimize(
  65                     errorOnLastNdays, c(0,7), simtype="endo")$minimum
  66             }
  67
  68             if (simtype == "endo")
  69             {
  70                 return (private$.predictShapeAux(data, fdays, today, horizon, local,
  71                     best_window_endo, "endo", TRUE))
  72             }
  73             if (simtype == "exo")
  74             {
  75                 return (private$.predictShapeAux(data, fdays, today, horizon, local,
  76                     best_window_exo, "exo", TRUE))
  77             }
  78             if (simtype == "mix")
  79             {
  80                 return(private$.predictShapeAux(data, fdays, today, horizon, local,
  81                     c(best_window_endo,best_window_exo), "mix", TRUE))
  82             }
  83         }
  84     ),
  85     private = list(
  86         # Precondition: "today" is full (no NAs)
  87         .predictShapeAux = function(data, fdays, today, horizon, local, window, simtype,
  88             final_call)
  89         {
  90             fdays_cut = fdays[ fdays < today ]
  91             if (length(fdays_cut) <= 1)
  92                 return (NA)
  93
  94             if (local)
  95             {
  96                 # Neighbors: days in "same season"; TODO: 60 == magic number...
  97                 fdays = getSimilarDaysIndices(today, data, limit=60, same_season=TRUE,
  98                     days_in=fdays_cut)
  99                 if (length(fdays) <= 1)
 100                     return (NA)
 101                 levelToday = data$getLevel(today)
 102                 distances = sapply(fdays, function(i) abs(data$getLevel(i)-levelToday))
 103                 #TODO: 2, 3, 5, 10 magic numbers here...
 104                 dist_thresh = 2
 105                 min_neighbs = min(3,length(fdays))
 106                 repeat
 107                 {
 108                     same_pollution = (distances <= dist_thresh)
 109                     nb_neighbs = sum(same_pollution)
 110                     if (nb_neighbs >= min_neighbs) #will eventually happen
 111                         break
 112                     dist_thresh = dist_thresh + 3
 113                 }
 114                 fdays = fdays[same_pollution]
 115                 max_neighbs = 10
 116                 if (nb_neighbs > max_neighbs)
 117                 {
 118                     # Keep only max_neighbs closest neighbors
 119                     fdays = fdays[
 120                         sort(distances[same_pollution],index.return=TRUE)$ix[1:max_neighbs] ]
 121                 }
 122                 if (length(fdays) == 1) #the other extreme...
 123                 {
 124                     if (final_call)
 125                     {
 126                         private$.params$weights <- 1
 127                         private$.params$indices <- fdays
 128                         private$.params$window <- 1
 129                     }
 130                     return ( data$getSerie(fdays[1])[1:horizon] ) #what else?!
 131                 }
 132             }
 133             else
 134                 fdays = fdays_cut #no conditioning
 135
 136             if (simtype != "exo")
 137             {
 138                 # Compute endogen similarities using given window
 139                 window_endo = ifelse(simtype=="mix", window[1], window)
 140
 141                 # Distances from last observed day to days in the past
 142                 serieToday = data$getSerie(today)
 143                 distances2 = sapply(fdays, function(i) {
 144                     delta = serieToday - data$getSerie(i)
 145                     mean(delta^2)
 146                 })
 147
 148                 sd_dist = sd(distances2)
 149                 if (sd_dist < .25 * sqrt(.Machine$double.eps))
 150                 {
 151 #                   warning("All computed distances are very close: stdev too small")
 152                     sd_dist = 1 #mostly for tests... FIXME:
 153                 }
 154                 simils_endo = exp(-distances2/(sd_dist*window_endo^2))
 155             }
 156
 157             if (simtype != "endo")
 158             {
 159                 # Compute exogen similarities using given window
 160                 h_exo = ifelse(simtype=="mix", window[2], window)
 161
 162                 M = matrix( nrow=1+length(fdays), ncol=1+length(data$getExo(today)) )
 163                 M[1,] = c( data$getLevel(today), as.double(data$getExo(today)) )
 164                 for (i in seq_along(fdays))
 165                     M[i+1,] = c( data$getLevel(fdays[i]), as.double(data$getExo(fdays[i])) )
 166
 167                 sigma = cov(M) #NOTE: robust covariance is way too slow
 168                 # TODO: 10 == magic number; more robust way == det, or always ginv()
 169                 sigma_inv =
 170                     if (length(fdays) > 10)
 171                         solve(sigma)
 172                     else
 173                         MASS::ginv(sigma)
 174
 175                 # Distances from last observed day to days in the past
 176                 distances2 = sapply(seq_along(fdays), function(i) {
 177                     delta = M[1,] - M[i+1,]
 178                     delta %*% sigma_inv %*% delta
 179                 })
 180
 181                 sd_dist = sd(distances2)
 182                 if (sd_dist < .25 * sqrt(.Machine$double.eps))
 183                 {
 184 #                   warning("All computed distances are very close: stdev too small")
 185                     sd_dist = 1 #mostly for tests... FIXME:
 186                 }
 187                 simils_exo = exp(-distances2/(sd_dist*window_exo^2))
 188             }
 189
 190             similarities =
 191                 if (simtype == "exo")
 192                     simils_exo
 193                 else if (simtype == "endo")
 194                     simils_endo
 195                 else #mix
 196                     simils_endo * simils_exo
 197             similarities = similarities / sum(similarities)
 198
 199             prediction = rep(0, horizon)
 200             for (i in seq_along(fdays))
 201                 prediction = prediction + similarities[i] * data$getSerie(fdays[i]+1)[1:horizon]
 202
 203             if (final_call)
 204             {
 205                 private$.params$weights <- similarities
 206                 private$.params$indices <- fdays
 207                 private$.params$window <-
 208                     if (simtype=="endo")
 209                         window_endo
 210                     else if (simtype=="exo")
 211                         window_exo
 212                     else #mix
 213                         c(window_endo,window_exo)
 214             }
 215
 216             return (prediction)
 217         }
 218     )
 219 )