| 1 | #' getNeighborsJumpPredict |
| 2 | #' |
| 3 | #' Apply optimized weights on gaps observed on selected neighbors. |
| 4 | #' This jump prediction method can only be used in conjunction with the Neighbors |
| 5 | #' Forecaster, because it makes use of the optimized parameters to re-apply the weights |
| 6 | #' on the jumps observed at days interfaces of the past neighbors. |
| 7 | #' |
| 8 | #' @inheritParams computeForecast |
| 9 | #' @inheritParams getZeroJumpPredict |
| 10 | #' |
| 11 | #' @aliases J_Neighbors |
| 12 | #' |
| 13 | getNeighborsJumpPredict = function(data, today, memory, horizon, params, ...) |
| 14 | { |
| 15 | first_day = max(1, today-memory) |
| 16 | filter = (params$indices >= first_day) |
| 17 | indices = params$indices[filter] |
| 18 | weights = params$weights[filter] |
| 19 | |
| 20 | if (any(is.na(weights) | is.na(indices))) |
| 21 | return (NA) |
| 22 | |
| 23 | gaps = sapply(indices, function(i) { |
| 24 | head( data$getSerie(i+1), 1) - tail( data$getSerie(i), 1) |
| 25 | }) |
| 26 | scal_product = weights * gaps |
| 27 | norm_fact = sum( weights[!is.na(scal_product)] ) |
| 28 | sum(scal_product, na.rm=TRUE) / norm_fact |
| 29 | } |