将多个对象强制转换为具有不同 start/end 日期的时间序列对象

Question

我正在学习本教程，使用扫描包对时间序列组执行整洁的时间序列预测。 Sweep 扩展了 broom 包以整理预测对象。

教程在这里： https://rdrr.io/cran/sweep/f/vignettes/SW01_Forecasting_Time_Series_Groups.Rmd

问题：我数据中的时间序列包含不同的长度和开始日期。在教程中，一个固定的开始被传递给 tk_ts() 因为每个时间序列都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：如何像上面的示例（以及在教程中）一样使用地图创建时间序列对象的列表列，但包括正确的开始日期和结束日期每个系列（每个系列都不同）

包：

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

可重现的示例数据：

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)

从这里我想应用一些函数来改变一个新的列表列，它包含来自 data.tbl 的相同数据，就像上面的例子（和教程中）强制转换为 ts 对象（为了与预测包一起使用）但每个系列的开始和结束日期都正确。

我想应用这样的东西：

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

但是问题是这只给出了 series_1 的正确开始日期。

如何使用每个系列的正确开始和结束日期来改变这个新的 ts 对象列表列？

谢谢

Answer 1

使用 format() 提取年份和月份作为 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 start/end 日期的时间序列对象

Coerce multiple objects to time series objects with different start/end dates

r

time-series

tidyverse

purrr

tidyquant