Question

我正在使用扫描包来执行本教程，以便对时间序列组执行整齐的时间序列预测。 Sweep将扫帚包扩展到整洁的预测对象。

这里的教程： https://rdrr.io/cran/sweep/f/vignettes/SW01_Forecasting_Time_Series_Groups.Rmd

问题：我的数据中的时间序列包含不同的长度和开始日期。在本教程中，固定的开始传递给tk_ts（），因为每个时间序列具有相同的开始和结束日期：

monthly_qty_by_cat2_ts <- monthly_qty_by_cat2_nest %>%
mutate(data.ts = map(.x       = data.tbl, 
                     .f       = tk_ts, 
                     select   = -order.month, 
                     start    = 2011, # <- see the fixed start date here
                     freq     = 12))

问题：如何使用上面的示例（以及教程中）创建时间序列对象的列表列但是包括每个系列的正确开始日期和结束日期（这是每个系列都不同）

的软件包：

library(tidyquant)
library(sweep)
library(timetk)
library(forecast)
library(tidyverse)

可重复的样本数据：

df <- structure(list(id = c("series_1", "series_1", "series_1", "series_1", 
"series_1", "series_1", "series_1", "series_1", "series_1", "series_1", 
"series_1", "series_1", "series_2", "series_2", "series_2", "series_2", 
"series_2", "series_2", "series_2", "series_2", "series_2", "series_2", 
"series_2", "series_2", "series_2", "series_2", "series_2", "series_2", 
"series_2", "series_2", "series_2", "series_2", "series_2", "series_2", 
"series_2", "series_2", "series_3", "series_3", "series_3", "series_3", 
"series_3", "series_3", "series_3", "series_3", "series_3", "series_3", 
"series_3", "series_3", "series_3", "series_3", "series_3", "series_3", 
"series_3", "series_3", "series_3", "series_3", "series_3", "series_3", 
"series_3", "series_3", "series_3", "series_3", "series_3", "series_3", 
"series_3", "series_3", "series_3", "series_3", "series_3", "series_3", 
"series_3", "series_3"), date = structure(c(10957, 10988, 11017, 
11048, 11078, 11109, 11139, 11170, 11201, 11231, 11262, 11292, 
13787, 13818, 13848, 13879, 13910, 13939, 13970, 14000, 14031, 
14061, 14092, 14123, 14153, 14184, 14214, 14245, 14276, 14304, 
14335, 14365, 14396, 14426, 14457, 14488, 15706, 15737, 15765, 
15796, 15826, 15857, 15887, 15918, 15949, 15979, 16010, 16040, 
16071, 16102, 16130, 16161, 16191, 16222, 16252, 16283, 16314, 
16344, 16375, 16405, 16436, 16467, 16495, 16526, 16556, 16587, 
16617, 16648, 16679, 16709, 16740, 16770), class = "Date"), value = c(0.526816892903298, 
0.0640646643005311, 0.569032567087561, 0.733993547270074, 0.742038151714951, 
0.273655793862417, 0.167404572479427, 0.766059899237007, 0.60176682821475, 
0.0769246644340456, 0.162491872673854, 0.323168716160581, 0.179594057612121, 
1.096650313586, 0.894524970557541, 1.55353087605909, 1.50662920810282, 
1.06641945429146, 1.95049989689142, 0.226111006457359, 0.644822218455374, 
0.998987099621445, 0.303691457025707, 0.782052680384368, 1.59218573896214, 
0.171859007328749, 1.9222901831381, 1.4127164632082, 0.919900813139975, 
1.93520273640752, 0.00968976970762014, 0.204170028213412, 1.90123205445707, 
1.05964627675712, 1.40747981145978, 0.476186634972692, 1.56826665904373, 
0.106335987104103, 2.7993093256373, 1.07078968570568, 0.668198951287195, 
0.584522894583642, 0.753677956061438, 2.76492932089604, 2.17496411106549, 
2.56561762047932, 0.586419345578179, 1.7261581714265, 1.38705582660623, 
0.708714888431132, 1.91359720285982, 1.85413848585449, 1.85429209470749, 
2.18856360157952, 1.00432092184201, 0.588805445702747, 2.95583719946444, 
0.382465981179848, 0.711439447710291, 1.24924974096939, 0.961857272777706, 
2.26519317110069, 1.10985011514276, 0.938654307508841, 0.985875837039202, 
1.13028976111673, 2.90536748478189, 0.795255574397743, 1.4741945641581, 
2.02167924796231, 1.2093570465222, 1.47486943169497)), .Names = c("id", 
"date", "value"), class = c("tbl_df", "tbl", "data.frame"), row.names = c(NA, 
-72L))

嵌套后：

df_nest <- df %>% group_by(id) %>% 
  nest(.key = data.tbl)

从这里开始，我想应用一些函数来改变一个新的列表列，该列包含来自data.tbl的相同数据，就像上面的例子（以及教程中）强制转换为ts对象一样（为了与之一起使用）预测包）但每个系列的正确开始和结束日期。

我想申请这样的话：

df_ts <- df_nest %>%
  mutate(data.ts = map(.x = data.tbl,
                       .f = tk_ts,
                       select = -date,
                       start = c(2000, 1), # <- Problem HERE
                       freq = 12))

但问题是这只会为series_1提供正确的开始日期。

如何使用每个系列的正确开始日期和结束日期来改变ts对象的新列表列？

由于

Answer 1

使用format()将年份和月份提取为start：

df_ts_2 <- df_nest %>%
  mutate(data.ts = map(.x = data.tbl,
                       .f = function(data) tk_ts(
                         data, 
                         select = -date, 
                         start = as.integer(c(format(data$date[1], "%Y"), format(data$date[1], "%m"))),
                         freq = 12
                       )))

print(df_ts_2$data.ts)

# [[1]]
#             Jan        Feb        Mar        Apr        May        Jun        Jul        Aug        Sep        Oct        Nov        Dec
# 2000 0.52681689 0.06406466 0.56903257 0.73399355 0.74203815 0.27365579 0.16740457 0.76605990 0.60176683 0.07692466 0.16249187 0.32316872
# 
# [[2]]
#             Jan        Feb        Mar        Apr        May        Jun        Jul        Aug        Sep        Oct        Nov        Dec
# 2007                                                                                                    0.17959406 1.09665031 0.89452497
# 2008 1.55353088 1.50662921 1.06641945 1.95049990 0.22611101 0.64482222 0.99898710 0.30369146 0.78205268 1.59218574 0.17185901 1.92229018
# 2009 1.41271646 0.91990081 1.93520274 0.00968977 0.20417003 1.90123205 1.05964628 1.40747981 0.47618663                                 
# 
# [[3]]
#            Jan       Feb       Mar       Apr       May       Jun       Jul       Aug       Sep       Oct       Nov       Dec
# 2013 1.5682667 0.1063360 2.7993093 1.0707897 0.6681990 0.5845229 0.7536780 2.7649293 2.1749641 2.5656176 0.5864193 1.7261582
# 2014 1.3870558 0.7087149 1.9135972 1.8541385 1.8542921 2.1885636 1.0043209 0.5888054 2.9558372 0.3824660 0.7114394 1.2492497
# 2015 0.9618573 2.2651932 1.1098501 0.9386543 0.9858758 1.1302898 2.9053675 0.7952556 1.4741946 2.0216792 1.2093570 1.4748694

将多个对象强制转换为具有不同开始/结束日期的时间序列对象

1 个答案: