R/ts_forecast.R
In yvesmauron/univariate-time-series-forecasting: Intelligent Algorithms to be Integrated in Business Intelligence solutions.
Defines functions
rx_sql_forunco
forecast_forunco
Documented in
forecast_forunco
rx_sql_forunco
#' Wrapper to forecast with RevoscaleR
#'
#' Create a wrapper around forunco that takes data from sql server or spark and
#' predicts the future timesteps with forunco logic.
#'
#' @param keys key of the time series, integer
#' @param data data of the time series passed by rxExecBy
#' @param params function parameters by rxExecBy
#'
#' @return forunco object
".rx_forecast" <- function(keys,
data,
h,
levels,
methods,
pp_methods,
point_combination,
pi_combination_upper,
pi_combination_lower,
pool_limit,
error_fun,
weight_fun,
val_h,
sov_only,
max_years,
val_min_years,
cv_min_years,
cv_max_samples,
allow_negatives,
...) {
# get data --> SELECT * FROM Target WHERE KeyCol = Key
inputData <- rxImport(inData = data)
# order data.frame by date
inputData <- inputData[order(as.Date(inputData$Date)),]
# data to date_int
date_int <- lubridate::decimal_date(as.Date(inputData$Date))
# create time series object
y <- ts(data = as.vector(inputData$Value), end = tail(date_int, 1), frequency = unique(inputData$Frequency))
# create
forunco <- forunco::forunco(
y = y,
h = h,
levels = levels,
methods = methods,
pp_methods = pp_methods,
point_combination = point_combination,
pi_combination_upper = pi_combination_upper,
pi_combination_lower = pi_combination_lower,
pool_limit = pool_limit,
error_fun = error_fun,
weight_fun = weight_fun,
val_h = val_h,
sov_only = sov_only,
max_years = max_years,
val_min_years = val_min_years,
cv_min_years = cv_min_years,
cv_max_samples = cv_max_samples,
allow_negatives = allow_negatives
)
return(forunco)
}
#' forunco function for SQL Server compute context
#'
#' Implements the forunco function for optimal parallelisation with machine learning services of sql server 2017
#'
#' @param connection_string mandatory: Connectionstring to the database
#' @param table mandatory: Table name that is the source for the forecasting objects
#' @param num_cores number of cores to be used, default \code{8}.
#' @param h number of horizons to predict.
#' @param levels Prediction interval levels. Optional, default \code{c(95)}.
#' @param methods Methods to be combined. Optional, default \code{auto_ets, auto_arima, auto_dotm}.
#' @param point_combination Point forecast combination operator. Optional, default \code{meidan}.
#' @param pi_combination_upper combination operator of the upper bound of the prediction intervals. Optional, default \code{max}.
#' @param pi_combination_lower combination operator for the lower bounds of the prediction intervals. Optional, default \code{min}.
#' @param pool_limit number of methods selected from the pool to reach the final forecasts. Optional, default \code{length(methods)}.
#' @param error_fun error function to determine the validation performance Optional, default \code{rmse}.
#' @param weight_fun weight function for calculating the definitive weights for combination. Optional, default \code{inverse}
#' @param val_h The horizon for the validation samples. Optional, default \code{h}.
#' @param sov_only Flag indicating whether only single origin validation should be considered. Optional, default \code{TRUE}.
#' @param max_years Maxmium of years to consider during model fitting. Optional, default \code{30}.
#' @param val_min_years Minimum years required to conduct single origin validation. Optional, default \code{4}.
#' @param cv_min_years Minimum years required to conduct cross-origin validation. Optional, default \code{5}.
#' @param cv_max_samples Maximum samples that should be considered during cross-validation, i.e. 3 indicates the algorithm validates from 3 origins. Optional, default \code{3}.
#' @param allow_negatives Flag indicating whether to allow negative values or not. Optional, default \code{FALSE}.
#' @param ...
#'
#' @return forecest data.frame
#' @export
#'
#' @examples
#' \dontrun{
#' connection_string <- "Server=LTYMA01\\QWERTZ;Database=FORECASTING_DATA;UID=ruser;PWD=ruser;"
#' table <- paste0("m3.vw_Micro")
#' result <- rx_sql_forunco(connection_string=connection_string, table=table)
#' # .id TSN date type value
#' # P1 1403 1994-03-02 Point 1468.48451
#' # P1 1403 1994-03-30 Point 1468.48451
#' # P1 1403 1994-04-30 Point 1468.48451
#' # P1 1403 1994-05-30 Point 1468.48451
#' # P1 1403 1994-06-30 Point 1468.48451
#' }
rx_sql_forunco <- function(connection_string,
table,
num_cores = 8,
h = 2,
levels = c(95),
methods = c('auto_ets', 'auto_arima', 'auto_thetaf'),
pp_methods = c('boxcox'),
point_combination = 'median',
pi_combination_upper = 'median',
pi_combination_lower = 'median',
pool_limit = length(methods),
error_fun = 'rmse',
weight_fun = 'inverse',
val_h = h,
sov_only = F,
max_years = 30,
val_min_years = 4,
cv_min_years = 5,
cv_max_samples = 3,
allow_negatives = F,
# tbd: additional parameters
# of sql server have
...) {
require(plyr)
require(dplyr)
if (!(RevoscaleR %in% installed.packages())) {
stop("This function can only run on SQL
ML Services or MRO Client with the Revoscaler package")
}
# define SQL Server Data source
sqlServerDataDS <- RxSqlServerData(table = table, connectionString = connection_string)
# Set SQL Server compute context with level of parallelism = 2
sqlServerCC <- RxInSqlServer(
connectionString = connection_string,
numTasks = 1,
packagesToLoad = c('forunco', 'dplyr')
)
# set most important options
rxOptions(numCoresToUse=num_cores)
# set compute context
rxSetComputeContext(sqlServerCC)
params = list(
h = h,
levels = levels,
methods = methods,
pp_methods = pp_methods,
point_combination = point_combination,
pi_combination_upper = pi_combination_upper,
pi_combination_lower = pi_combination_lower,
pool_limit = pool_limit,
error_fun = error_fun,
weight_fun = weight_fun,
val_h = val_h,
sov_only = sov_only,
max_years = max_years,
val_min_years = val_min_years,
cv_min_years = cv_min_years,
cv_max_samples = cv_max_samples,
allow_negatives = allow_negatives
)
# execute prediction function
qlServerCCResults <- rxExecBy(
inData = sqlServerDataDS,
keys = c("TSN"),
func = .rx_forecast,
funcParams = params
)
res <- lapply(qlServerCCResults, function(x) {
data.frame("TSN"=x$keys[[1]][[1]], x$result$preds)
})
# return result
res <- plyr::ldply(res, rbind) %>%
dplyr::mutate(date = as.character(date))
return(res)
}
#' Forunco function for batch forecasting in R
#'
#' Implements the forunco function for optimal parallelisation with machine learning services of sql server 2017
#'
#' @param ts_col list of time series objects
#' @param h number of horizons to predict
#' @param num_cores number of cores to be used; default: NULL (all cores)
#' @param num_cores_ignore number of cores to be ignored for e.g. OS; default: 1
#' @param prog_bar boolean, whether or not a progress bar should be displyed
#' @param levels Prediction interval levels. Optional, default \code{c(95)}.
#' @param methods Methods to be combined. Optional, default \code{auto_ets, auto_arima, auto_dotm}.
#' @param point_combination Point forecast combination operator. Optional, default \code{meidan}.
#' @param pi_combination_upper combination operator of the upper bound of the prediction intervals. Optional, default \code{max}.
#' @param pi_combination_lower combination operator for the lower bounds of the prediction intervals. Optional, default \code{min}.
#' @param pool_limit number of methods selected from the pool to reach the final forecasts. Optional, default \code{length(methods)}.
#' @param error_fun error function to determine the validation performance Optional, default \code{rmse}.
#' @param weight_fun weight function for calculating the definitive weights for combination. Optional, default \code{inverse}
#' @param val_h The horizon for the validation samples. Optional, default \code{h}.
#' @param sov_only Flag indicating whether only single origin validation should be considered. Optional, default \code{TRUE}.
#' @param max_years Maxmium of years to consider during model fitting. Optional, default \code{30}.
#' @param val_min_years Minimum years required to conduct single origin validation. Optional, default \code{4}.
#' @param cv_min_years Minimum years required to conduct cross-origin validation. Optional, default \code{5}.
#' @param cv_max_samples Maximum samples that should be considered during cross-validation, i.e. 3 indicates the algorithm validates from 3 origins. Optional, default \code{3}.
#' @param allow_negatives Flag indicating whether to allow negative values or not. Optional, default \code{FALSE}.
#' @param ...
#'
#' @return list with mean, pis and data_frame containing the predicted values
#' @export
#'
#' @examples
#' \dontrun{
#' library(forunco)
#' library(Mcomp)
#' m3 <- M3[2000:2025]
#' #a time series colllection has n elements, each of which is a time series
#' # object.
#' ts_col <- lapply(m3, function(x) {x$x})
#'
#' result <- forecast_forunco(ts_col)
#' preds[[1]]
#' $mean
#' [1] 5054.439 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170
#'
#' $upper
#' [1] 6348.204 6883.349 7294.081 7640.378 7945.488 8221.338 8475.014 8711.134 8932.906 9142.665 9342.174 9532.805
#'
#' $lower
#' [1] 3760.6740 3149.1485 2453.7666 2132.0499 2054.5814 1887.5394 1633.8633 1397.7433 1175.9717 894.6316 766.7032 576.0728
#'
#' $preds
#' # A tibble: 36 X 3
#' date type value
#' <date> <chr> <dbl>
#' 1989-07-02 Point 5054.439
#' 1989-08-02 Point 5048.170
#' 1989-09-02 Point 5048.170
#' 1989-10-02 Point 5048.170
#' 1989-11-02 Point 5048.170
#' 1989-12-02 Point 5048.170
#' 1990-01-02 Point 5048.170
#' 1990-02-02 Point 5048.170
#' 1990-03-02 Point 5048.170
#' 1990-04-02 Point 5048.170
# ... with 26 more rows
#' }
forecast_forunco <- function(ts_col,
h = 12,
num_cores = NULL,
num_cores_ignore = 1,
prog_bar = T,
levels = c(95),
methods = c('auto_ets', 'auto_arima', 'auto_dotm'),
point_combination = 'median',
pi_combination_upper = 'median',
pi_combination_lower = 'median',
pool_limit = length(methods),
error_fun = 'rmse',
weight_fun = 'inverse',
val_h = h,
sov_only = F,
max_years = 30,
val_min_years = 4,
cv_min_years = 5,
cv_max_samples = 3,
allow_negatives = F,
# tbd: additional parameters
# of sql server
...) {
# load packages
#require(forunco)
require(foreach)
require(doSNOW)
require(parallel)
require(progress)
# create progress bar object
if (prog_bar) {
# progress bar object
pb <- progress_bar$new(
format = " [:bar] :percent eta: :eta",
total = length(ts_col),
clear = F,
force = T
)
prop_progress <- function() { pb$tick() }
} else {
prop_progress <- function() { NULL }
}
# set num_cores if it is not set by the user
if (is.null(num_cores))
num_cores <- detectCores()
cl <- makeCluster(num_cores - num_cores_ignore, type = "SOCK")
registerDoSNOW(cl)
clusterEvalQ(cl, {
library(forunco);
library(dplyr);
library(forecast)
})
#clusterExport(cl, ls())
#
tc <- foreach (i = seq.int(length.out = length(ts_col)), .options.snow = list(progress=prop_progress)) %dopar% {
ts <- ts_col[[i]]
forunco(y = ts,
h = h,
levels = levels,
methods = methods,
point_combination = point_combination,
pi_combination_upper = pi_combination_upper,
pi_combination_lower = pi_combination_lower,
pool_limit = pool_limit,
error_fun = error_fun,
weight_fun = weight_fun,
val_h = val_h,
sov_only = sov_only,
max_years = max_years,
val_min_years = val_min_years,
cv_min_years = cv_min_years,
cv_max_samples = cv_max_samples,
allow_negatives = allow_negatives
)
}
stopCluster(cl)
return(tc)
}
yvesmauron/univariate-time-series-forecasting documentation built on March 2, 2020, 12:20 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions rx_sql_forunco forecast_forunco
Documented in forecast_forunco rx_sql_forunco
#' Wrapper to forecast with RevoscaleR
#'
#' Create a wrapper around forunco that takes data from sql server or spark and
#' predicts the future timesteps with forunco logic.
#'
#' @param keys key of the time series, integer
#' @param data data of the time series passed by rxExecBy
#' @param params function parameters by rxExecBy
#'
#' @return forunco object
".rx_forecast" <- function(keys,
data,
h,
levels,
methods,
pp_methods,
point_combination,
pi_combination_upper,
pi_combination_lower,
pool_limit,
error_fun,
weight_fun,
val_h,
sov_only,
max_years,
val_min_years,
cv_min_years,
cv_max_samples,
allow_negatives,
...) {
# get data --> SELECT * FROM Target WHERE KeyCol = Key
inputData <- rxImport(inData = data)
# order data.frame by date
inputData <- inputData[order(as.Date(inputData$Date)),]
# data to date_int
date_int <- lubridate::decimal_date(as.Date(inputData$Date))
# create time series object
y <- ts(data = as.vector(inputData$Value), end = tail(date_int, 1), frequency = unique(inputData$Frequency))
# create
forunco <- forunco::forunco(
y = y,
h = h,
levels = levels,
methods = methods,
pp_methods = pp_methods,
point_combination = point_combination,
pi_combination_upper = pi_combination_upper,
pi_combination_lower = pi_combination_lower,
pool_limit = pool_limit,
error_fun = error_fun,
weight_fun = weight_fun,
val_h = val_h,
sov_only = sov_only,
max_years = max_years,
val_min_years = val_min_years,
cv_min_years = cv_min_years,
cv_max_samples = cv_max_samples,
allow_negatives = allow_negatives
)
return(forunco)
}
#' forunco function for SQL Server compute context
#'
#' Implements the forunco function for optimal parallelisation with machine learning services of sql server 2017
#'
#' @param connection_string mandatory: Connectionstring to the database
#' @param table mandatory: Table name that is the source for the forecasting objects
#' @param num_cores number of cores to be used, default \code{8}.
#' @param h number of horizons to predict.
#' @param levels Prediction interval levels. Optional, default \code{c(95)}.
#' @param methods Methods to be combined. Optional, default \code{auto_ets, auto_arima, auto_dotm}.
#' @param point_combination Point forecast combination operator. Optional, default \code{meidan}.
#' @param pi_combination_upper combination operator of the upper bound of the prediction intervals. Optional, default \code{max}.
#' @param pi_combination_lower combination operator for the lower bounds of the prediction intervals. Optional, default \code{min}.
#' @param pool_limit number of methods selected from the pool to reach the final forecasts. Optional, default \code{length(methods)}.
#' @param error_fun error function to determine the validation performance Optional, default \code{rmse}.
#' @param weight_fun weight function for calculating the definitive weights for combination. Optional, default \code{inverse}
#' @param val_h The horizon for the validation samples. Optional, default \code{h}.
#' @param sov_only Flag indicating whether only single origin validation should be considered. Optional, default \code{TRUE}.
#' @param max_years Maxmium of years to consider during model fitting. Optional, default \code{30}.
#' @param val_min_years Minimum years required to conduct single origin validation. Optional, default \code{4}.
#' @param cv_min_years Minimum years required to conduct cross-origin validation. Optional, default \code{5}.
#' @param cv_max_samples Maximum samples that should be considered during cross-validation, i.e. 3 indicates the algorithm validates from 3 origins. Optional, default \code{3}.
#' @param allow_negatives Flag indicating whether to allow negative values or not. Optional, default \code{FALSE}.
#' @param ...
#'
#' @return forecest data.frame
#' @export
#'
#' @examples
#' \dontrun{
#' connection_string <- "Server=LTYMA01\\QWERTZ;Database=FORECASTING_DATA;UID=ruser;PWD=ruser;"
#' table <- paste0("m3.vw_Micro")
#' result <- rx_sql_forunco(connection_string=connection_string, table=table)
#' # .id TSN date type value
#' # P1 1403 1994-03-02 Point 1468.48451
#' # P1 1403 1994-03-30 Point 1468.48451
#' # P1 1403 1994-04-30 Point 1468.48451
#' # P1 1403 1994-05-30 Point 1468.48451
#' # P1 1403 1994-06-30 Point 1468.48451
#' }
rx_sql_forunco <- function(connection_string,
table,
num_cores = 8,
h = 2,
levels = c(95),
methods = c('auto_ets', 'auto_arima', 'auto_thetaf'),
pp_methods = c('boxcox'),
point_combination = 'median',
pi_combination_upper = 'median',
pi_combination_lower = 'median',
pool_limit = length(methods),
error_fun = 'rmse',
weight_fun = 'inverse',
val_h = h,
sov_only = F,
max_years = 30,
val_min_years = 4,
cv_min_years = 5,
cv_max_samples = 3,
allow_negatives = F,
# tbd: additional parameters
# of sql server have
...) {
require(plyr)
require(dplyr)
if (!(RevoscaleR %in% installed.packages())) {
stop("This function can only run on SQL
ML Services or MRO Client with the Revoscaler package")
}
# define SQL Server Data source
sqlServerDataDS <- RxSqlServerData(table = table, connectionString = connection_string)
# Set SQL Server compute context with level of parallelism = 2
sqlServerCC <- RxInSqlServer(
connectionString = connection_string,
numTasks = 1,
packagesToLoad = c('forunco', 'dplyr')
)
# set most important options
rxOptions(numCoresToUse=num_cores)
# set compute context
rxSetComputeContext(sqlServerCC)
params = list(
h = h,
levels = levels,
methods = methods,
pp_methods = pp_methods,
point_combination = point_combination,
pi_combination_upper = pi_combination_upper,
pi_combination_lower = pi_combination_lower,
pool_limit = pool_limit,
error_fun = error_fun,
weight_fun = weight_fun,
val_h = val_h,
sov_only = sov_only,
max_years = max_years,
val_min_years = val_min_years,
cv_min_years = cv_min_years,
cv_max_samples = cv_max_samples,
allow_negatives = allow_negatives
)
# execute prediction function
qlServerCCResults <- rxExecBy(
inData = sqlServerDataDS,
keys = c("TSN"),
func = .rx_forecast,
funcParams = params
)
res <- lapply(qlServerCCResults, function(x) {
data.frame("TSN"=x$keys[[1]][[1]], x$result$preds)
})
# return result
res <- plyr::ldply(res, rbind) %>%
dplyr::mutate(date = as.character(date))
return(res)
}
#' Forunco function for batch forecasting in R
#'
#' Implements the forunco function for optimal parallelisation with machine learning services of sql server 2017
#'
#' @param ts_col list of time series objects
#' @param h number of horizons to predict
#' @param num_cores number of cores to be used; default: NULL (all cores)
#' @param num_cores_ignore number of cores to be ignored for e.g. OS; default: 1
#' @param prog_bar boolean, whether or not a progress bar should be displyed
#' @param levels Prediction interval levels. Optional, default \code{c(95)}.
#' @param methods Methods to be combined. Optional, default \code{auto_ets, auto_arima, auto_dotm}.
#' @param point_combination Point forecast combination operator. Optional, default \code{meidan}.
#' @param pi_combination_upper combination operator of the upper bound of the prediction intervals. Optional, default \code{max}.
#' @param pi_combination_lower combination operator for the lower bounds of the prediction intervals. Optional, default \code{min}.
#' @param pool_limit number of methods selected from the pool to reach the final forecasts. Optional, default \code{length(methods)}.
#' @param error_fun error function to determine the validation performance Optional, default \code{rmse}.
#' @param weight_fun weight function for calculating the definitive weights for combination. Optional, default \code{inverse}
#' @param val_h The horizon for the validation samples. Optional, default \code{h}.
#' @param sov_only Flag indicating whether only single origin validation should be considered. Optional, default \code{TRUE}.
#' @param max_years Maxmium of years to consider during model fitting. Optional, default \code{30}.
#' @param val_min_years Minimum years required to conduct single origin validation. Optional, default \code{4}.
#' @param cv_min_years Minimum years required to conduct cross-origin validation. Optional, default \code{5}.
#' @param cv_max_samples Maximum samples that should be considered during cross-validation, i.e. 3 indicates the algorithm validates from 3 origins. Optional, default \code{3}.
#' @param allow_negatives Flag indicating whether to allow negative values or not. Optional, default \code{FALSE}.
#' @param ...
#'
#' @return list with mean, pis and data_frame containing the predicted values
#' @export
#'
#' @examples
#' \dontrun{
#' library(forunco)
#' library(Mcomp)
#' m3 <- M3[2000:2025]
#' #a time series colllection has n elements, each of which is a time series
#' # object.
#' ts_col <- lapply(m3, function(x) {x$x})
#'
#' result <- forecast_forunco(ts_col)
#' preds[[1]]
#' $mean
#' [1] 5054.439 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170 5048.170
#'
#' $upper
#' [1] 6348.204 6883.349 7294.081 7640.378 7945.488 8221.338 8475.014 8711.134 8932.906 9142.665 9342.174 9532.805
#'
#' $lower
#' [1] 3760.6740 3149.1485 2453.7666 2132.0499 2054.5814 1887.5394 1633.8633 1397.7433 1175.9717 894.6316 766.7032 576.0728
#'
#' $preds
#' # A tibble: 36 X 3
#' date type value
#' <date> <chr> <dbl>
#' 1989-07-02 Point 5054.439
#' 1989-08-02 Point 5048.170
#' 1989-09-02 Point 5048.170
#' 1989-10-02 Point 5048.170
#' 1989-11-02 Point 5048.170
#' 1989-12-02 Point 5048.170
#' 1990-01-02 Point 5048.170
#' 1990-02-02 Point 5048.170
#' 1990-03-02 Point 5048.170
#' 1990-04-02 Point 5048.170
# ... with 26 more rows
#' }
forecast_forunco <- function(ts_col,
h = 12,
num_cores = NULL,
num_cores_ignore = 1,
prog_bar = T,
levels = c(95),
methods = c('auto_ets', 'auto_arima', 'auto_dotm'),
point_combination = 'median',
pi_combination_upper = 'median',
pi_combination_lower = 'median',
pool_limit = length(methods),
error_fun = 'rmse',
weight_fun = 'inverse',
val_h = h,
sov_only = F,
max_years = 30,
val_min_years = 4,
cv_min_years = 5,
cv_max_samples = 3,
allow_negatives = F,
# tbd: additional parameters
# of sql server
...) {
# load packages
#require(forunco)
require(foreach)
require(doSNOW)
require(parallel)
require(progress)
# create progress bar object
if (prog_bar) {
# progress bar object
pb <- progress_bar$new(
format = " [:bar] :percent eta: :eta",
total = length(ts_col),
clear = F,
force = T
)
prop_progress <- function() { pb$tick() }
} else {
prop_progress <- function() { NULL }
}
# set num_cores if it is not set by the user
if (is.null(num_cores))
num_cores <- detectCores()
cl <- makeCluster(num_cores - num_cores_ignore, type = "SOCK")
registerDoSNOW(cl)
clusterEvalQ(cl, {
library(forunco);
library(dplyr);
library(forecast)
})
#clusterExport(cl, ls())
#
tc <- foreach (i = seq.int(length.out = length(ts_col)), .options.snow = list(progress=prop_progress)) %dopar% {
ts <- ts_col[[i]]
forunco(y = ts,
h = h,
levels = levels,
methods = methods,
point_combination = point_combination,
pi_combination_upper = pi_combination_upper,
pi_combination_lower = pi_combination_lower,
pool_limit = pool_limit,
error_fun = error_fun,
weight_fun = weight_fun,
val_h = val_h,
sov_only = sov_only,
max_years = max_years,
val_min_years = val_min_years,
cv_min_years = cv_min_years,
cv_max_samples = cv_max_samples,
allow_negatives = allow_negatives
)
}
stopCluster(cl)
return(tc)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.