X-Git-Url: https://git.auder.net/?p=talweg.git;a=blobdiff_plain;f=pkg%2FR%2FF_Neighbors.R;h=9ba72b8f308fcf00ef43e99c5e968c731eac1fbe;hp=c55291aa996838a2a02535725f72c4fb8709b922;hb=445e7bbc18aa739ec0b3caba4d8710a9d9e1a43c;hpb=21c70378d89863afeb124d86989f7f956e280808 diff --git a/pkg/R/F_Neighbors.R b/pkg/R/F_Neighbors.R index c55291a..9ba72b8 100644 --- a/pkg/R/F_Neighbors.R +++ b/pkg/R/F_Neighbors.R @@ -22,8 +22,8 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", fdays = getNoNA2(data, max(today-memory,1), today-1) # Get optional args - local = ifelse(hasArg("local"), list(...)$local, FALSE) #same level + season? - simtype = ifelse(hasArg("simtype"), list(...)$simtype, "mix") #or "endo", or "exo" + local = ifelse(hasArg("local"), list(...)$local, TRUE) #same level + season? + simtype = ifelse(hasArg("simtype"), list(...)$simtype, "none") #or "endo", or "exo" if (hasArg("window")) { return ( private$.predictShapeAux(data, @@ -34,7 +34,7 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", cv_days = getSimilarDaysIndices(today, data, limit=20, same_season=FALSE, days_in=fdays) - # Optimize h : h |--> sum of prediction errors on last 45 "similar" days + # Optimize h : h |--> sum of prediction errors on last N "similar" days errorOnLastNdays = function(window, simtype) { error = 0 @@ -54,6 +54,7 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", return (error / nb_jours) } + # TODO: 7 == magic number if (simtype != "endo") { best_window_exo = optimize( @@ -100,9 +101,9 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", return (NA) levelToday = data$getLevel(today) distances = sapply(fdays, function(i) abs(data$getLevel(i)-levelToday)) - #TODO: 2, 3, 5, 10 magic numbers here... + #TODO: 2, 10, 3, 12 magic numbers here... dist_thresh = 2 - min_neighbs = min(3,length(fdays)) + min_neighbs = min(10,length(fdays)) repeat { same_pollution = (distances <= dist_thresh) @@ -112,7 +113,7 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", dist_thresh = dist_thresh + 3 } fdays = fdays[same_pollution] - max_neighbs = 10 + max_neighbs = 12 if (nb_neighbs > max_neighbs) { # Keep only max_neighbs closest neighbors @@ -133,7 +134,7 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", else fdays = fdays_cut #no conditioning - if (simtype != "exo") + if (simtype == "endo" || simtype == "mix") { # Compute endogen similarities using given window window_endo = ifelse(simtype=="mix", window[1], window) @@ -154,10 +155,10 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", simils_endo = exp(-distances2/(sd_dist*window_endo^2)) } - if (simtype != "endo") + if (simtype == "exo" || simtype == "mix") { # Compute exogen similarities using given window - h_exo = ifelse(simtype=="mix", window[2], window) + window_exo = ifelse(simtype=="mix", window[2], window) M = matrix( nrow=1+length(fdays), ncol=1+length(data$getExo(today)) ) M[1,] = c( data$getLevel(today), as.double(data$getExo(today)) ) @@ -192,8 +193,10 @@ NeighborsForecaster = R6::R6Class("NeighborsForecaster", simils_exo else if (simtype == "endo") simils_endo - else #mix + else if (simtype == "mix") simils_endo * simils_exo + else #none + rep(1, length(fdays)) similarities = similarities / sum(similarities) prediction = rep(0, horizon)