healthyverse_tsa
Time Series Analysis, Modeling and Forecasting of the Healthyverse
Packages Steven P. Sanderson II, MPH - Date: 2026-01-29
Introduction
This analysis follows a Nested Modeltime Workflow from modeltime
along with using the NNS package. I use this to monitor the
downloads of all of my packages:
Get Data
glimpse(downloads_tbl)
Rows: 166,986
Columns: 11
$ date <date> 2020-11-23, 2020-11-23, 2020-11-23, 2020-11-23, 2020-11-23,…
$ time <Period> 15H 36M 55S, 11H 26M 39S, 23H 34M 44S, 18H 39M 32S, 9H 0M…
$ date_time <dttm> 2020-11-23 15:36:55, 2020-11-23 11:26:39, 2020-11-23 23:34:…
$ size <int> 4858294, 4858294, 4858301, 4858295, 361, 4863722, 4864794, 4…
$ r_version <chr> NA, "4.0.3", "3.5.3", "3.5.2", NA, NA, NA, NA, NA, NA, NA, N…
$ r_arch <chr> NA, "x86_64", "x86_64", "x86_64", NA, NA, NA, NA, NA, NA, NA…
$ r_os <chr> NA, "mingw32", "mingw32", "linux-gnu", NA, NA, NA, NA, NA, N…
$ package <chr> "healthyR.data", "healthyR.data", "healthyR.data", "healthyR…
$ version <chr> "1.0.0", "1.0.0", "1.0.0", "1.0.0", "1.0.0", "1.0.0", "1.0.0…
$ country <chr> "US", "US", "US", "GB", "US", "US", "DE", "HK", "JP", "US", …
$ ip_id <int> 2069, 2804, 78827, 27595, 90474, 90474, 42435, 74, 7655, 638…
The last day in the data set is 2026-01-27 23:29:15, the file was birthed on: 2025-10-31 10:47:59.603742, and at report knit time is 2120.69 hours old. Happy analyzing!
Now that we have our data lets take a look at it using the skimr
package.
skim(downloads_tbl)
| Name | downloads_tbl |
| Number of rows | 166986 |
| Number of columns | 11 |
| _______________________ | |
| Column type frequency: | |
| character | 6 |
| Date | 1 |
| numeric | 2 |
| POSIXct | 1 |
| Timespan | 1 |
| ________________________ | |
| Group variables | None |
Data summary
Variable type: character
| skim_variable | n_missing | complete_rate | min | max | empty | n_unique | whitespace |
|---|---|---|---|---|---|---|---|
| r_version | 123015 | 0.26 | 5 | 7 | 0 | 50 | 0 |
| r_arch | 123015 | 0.26 | 1 | 7 | 0 | 6 | 0 |
| r_os | 123015 | 0.26 | 7 | 19 | 0 | 24 | 0 |
| package | 0 | 1.00 | 7 | 13 | 0 | 8 | 0 |
| version | 0 | 1.00 | 5 | 17 | 0 | 63 | 0 |
| country | 15636 | 0.91 | 2 | 2 | 0 | 166 | 0 |
Variable type: Date
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| date | 0 | 1 | 2020-11-23 | 2026-01-27 | 2023-11-30 | 1885 |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| size | 0 | 1 | 1124808.46 | 1484738.57 | 355 | 35495 | 313210 | 2347811 | 5677952 | ▇▁▂▁▁ |
| ip_id | 0 | 1 | 11207.45 | 21831.19 | 1 | 223 | 2793 | 11729 | 299146 | ▇▁▁▁▁ |
Variable type: POSIXct
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| date_time | 0 | 1 | 2020-11-23 09:00:41 | 2026-01-27 23:29:15 | 2023-11-30 02:14:37 | 105656 |
Variable type: Timespan
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| time | 0 | 1 | 0 | 59 | 52 | 60 |
We can see that the following columns are missing a lot of data and for
us are most likely not useful anyways, so we will drop them
c(r_version, r_arch, r_os)
Plots
Now lets take a look at a time-series plot of the total daily downloads by package. We will use a log scale and place a vertical line at each version release for each package.
[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
Now lets take a look at some time series decomposition graphs.
[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
Seasonal Diagnostics:
[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
ACF and PACF Diagnostics:
[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
Feature Engineering
Now that we have our basic data and a shot of what it looks like, let’s
add some features to our data which can be very helpful in modeling.
Lets start by making a tibble that is aggregated by the day and
package, as we are going to be interested in forecasting the next 4
weeks or 28 days for each package. First lets get our base data.
Call:
stats::lm(formula = .formula, data = df)
Residuals:
Min 1Q Median 3Q Max
-148.51 -37.04 -11.51 27.12 823.78
Coefficients:
Estimate Std. Error
(Intercept) -1.648e+02 5.758e+01
date 1.028e-02 3.048e-03
lag(value, 1) 1.028e-01 2.289e-02
lag(value, 7) 9.103e-02 2.370e-02
lag(value, 14) 7.741e-02 2.365e-02
lag(value, 21) 8.329e-02 2.372e-02
lag(value, 28) 6.242e-02 2.365e-02
lag(value, 35) 5.097e-02 2.365e-02
lag(value, 42) 6.777e-02 2.376e-02
lag(value, 49) 6.724e-02 2.369e-02
month(date, label = TRUE).L -8.960e+00 4.845e+00
month(date, label = TRUE).Q -9.597e-01 4.757e+00
month(date, label = TRUE).C -1.458e+01 4.803e+00
month(date, label = TRUE)^4 -7.292e+00 4.859e+00
month(date, label = TRUE)^5 -5.862e+00 4.851e+00
month(date, label = TRUE)^6 6.876e-01 4.896e+00
month(date, label = TRUE)^7 -4.111e+00 4.844e+00
month(date, label = TRUE)^8 -4.250e+00 4.823e+00
month(date, label = TRUE)^9 2.882e+00 4.837e+00
month(date, label = TRUE)^10 9.234e-01 4.853e+00
month(date, label = TRUE)^11 -4.123e+00 4.840e+00
fourier_vec(date, type = "sin", K = 1, period = 7) -1.101e+01 2.176e+00
fourier_vec(date, type = "cos", K = 1, period = 7) 7.265e+00 2.250e+00
t value Pr(>|t|)
(Intercept) -2.863 0.004245 **
date 3.372 0.000763 ***
lag(value, 1) 4.489 7.59e-06 ***
lag(value, 7) 3.842 0.000126 ***
lag(value, 14) 3.274 0.001082 **
lag(value, 21) 3.511 0.000458 ***
lag(value, 28) 2.639 0.008386 **
lag(value, 35) 2.155 0.031319 *
lag(value, 42) 2.852 0.004396 **
lag(value, 49) 2.839 0.004580 **
month(date, label = TRUE).L -1.849 0.064562 .
month(date, label = TRUE).Q -0.202 0.840156
month(date, label = TRUE).C -3.035 0.002441 **
month(date, label = TRUE)^4 -1.501 0.133599
month(date, label = TRUE)^5 -1.208 0.227034
month(date, label = TRUE)^6 0.140 0.888316
month(date, label = TRUE)^7 -0.849 0.396219
month(date, label = TRUE)^8 -0.881 0.378286
month(date, label = TRUE)^9 0.596 0.551333
month(date, label = TRUE)^10 0.190 0.849121
month(date, label = TRUE)^11 -0.852 0.394392
fourier_vec(date, type = "sin", K = 1, period = 7) -5.059 4.65e-07 ***
fourier_vec(date, type = "cos", K = 1, period = 7) 3.229 0.001263 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 59.37 on 1813 degrees of freedom
(49 observations deleted due to missingness)
Multiple R-squared: 0.2225, Adjusted R-squared: 0.2131
F-statistic: 23.58 on 22 and 1813 DF, p-value: < 2.2e-16
NNS Forecasting
This is something I have been wanting to try for a while. The NNS
package is a great package for forecasting time series data.
library(NNS)
data_list <- base_data |>
select(package, value) |>
group_split(package)
data_list |>
imap(
\(x, idx) {
obj <- x
x <- obj |> pull(value) |> tail(7*52)
train_set_size <- length(x) - 56
pkg <- obj |> pluck(1) |> unique()
# sf <- NNS.seas(x, modulo = 7, plot = FALSE)$periods
seas <- t(
sapply(
1:25,
function(i) c(
i,
sqrt(
mean((
NNS.ARMA(x,
h = 28,
training.set = train_set_size,
method = "lin",
seasonal.factor = i,
plot=FALSE
) - tail(x, 28)) ^ 2)))
)
)
colnames(seas) <- c("Period", "RMSE")
sf <- seas[which.min(seas[, 2]), 1]
cat(paste0("Package: ", pkg, "\n"))
NNS.ARMA.optim(
variable = x,
h = 28,
training.set = train_set_size,
#seasonal.factor = seq(12, 60, 7),
seasonal.factor = sf,
pred.int = 0.95,
plot = TRUE
)
title(
sub = paste0("\n",
"Package: ", pkg, " - NNS Optimization")
)
}
)
Package: healthyR
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 25 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 5.05681176423358"
[1] "BEST method = 'lin' PATH MEMBER = c( 25 )"
[1] "BEST lin OBJECTIVE FUNCTION = 5.05681176423358"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 25 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 6.12043640118858"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 25 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 6.12043640118858"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 25 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 4.70439047238501"
[1] "BEST method = 'both' PATH MEMBER = c( 25 )"
[1] "BEST both OBJECTIVE FUNCTION = 4.70439047238501"
Package: healthyR.ai
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 3 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 101.983050461472"
[1] "BEST method = 'lin' PATH MEMBER = c( 3 )"
[1] "BEST lin OBJECTIVE FUNCTION = 101.983050461472"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 3 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 10.3903729485468"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 3 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 10.3903729485468"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 3 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 9.91729626355765"
[1] "BEST method = 'both' PATH MEMBER = c( 3 )"
[1] "BEST both OBJECTIVE FUNCTION = 9.91729626355765"
Package: healthyR.data
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 20 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 12.7508897683222"
[1] "BEST method = 'lin' PATH MEMBER = c( 20 )"
[1] "BEST lin OBJECTIVE FUNCTION = 12.7508897683222"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 20 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 5.89455872851956"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 20 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 5.89455872851956"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 20 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 7.06643722643935"
[1] "BEST method = 'both' PATH MEMBER = c( 20 )"
[1] "BEST both OBJECTIVE FUNCTION = 7.06643722643935"
Package: healthyR.ts
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 11 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 26.6078901632946"
[1] "BEST method = 'lin' PATH MEMBER = c( 11 )"
[1] "BEST lin OBJECTIVE FUNCTION = 26.6078901632946"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 11 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 11.5754263945389"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 11 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 11.5754263945389"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 11 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 17.5896000127571"
[1] "BEST method = 'both' PATH MEMBER = c( 11 )"
[1] "BEST both OBJECTIVE FUNCTION = 17.5896000127571"
Package: healthyverse
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 3 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 61.9756592532723"
[1] "BEST method = 'lin' PATH MEMBER = c( 3 )"
[1] "BEST lin OBJECTIVE FUNCTION = 61.9756592532723"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 3 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 15.3999828610111"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 3 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 15.3999828610111"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 3 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 33.9948068539943"
[1] "BEST method = 'both' PATH MEMBER = c( 3 )"
[1] "BEST both OBJECTIVE FUNCTION = 33.9948068539943"
Package: RandomWalker
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 16 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 14.9179587262231"
[1] "BEST method = 'lin' PATH MEMBER = c( 16 )"
[1] "BEST lin OBJECTIVE FUNCTION = 14.9179587262231"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 16 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 4.60571522181754"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 16 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 4.60571522181754"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 16 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 6.2682288509071"
[1] "BEST method = 'both' PATH MEMBER = c( 16 )"
[1] "BEST both OBJECTIVE FUNCTION = 6.2682288509071"
Package: tidyAML
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 8 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 11.8387988264058"
[1] "BEST method = 'lin' PATH MEMBER = c( 8 )"
[1] "BEST lin OBJECTIVE FUNCTION = 11.8387988264058"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 8 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 11.5399288115919"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 8 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 11.5399288115919"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 8 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 10.5275570095888"
[1] "BEST method = 'both' PATH MEMBER = c( 8 )"
[1] "BEST both OBJECTIVE FUNCTION = 10.5275570095888"
Package: TidyDensity
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 16 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 19.3847801917076"
[1] "BEST method = 'lin' PATH MEMBER = c( 16 )"
[1] "BEST lin OBJECTIVE FUNCTION = 19.3847801917076"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 16 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 3.42317707972579"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 16 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 3.42317707972579"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 16 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 5.81528551130106"
[1] "BEST method = 'both' PATH MEMBER = c( 16 )"
[1] "BEST both OBJECTIVE FUNCTION = 5.81528551130106"
[[1]]
NULL
[[2]]
NULL
[[3]]
NULL
[[4]]
NULL
[[5]]
NULL
[[6]]
NULL
[[7]]
NULL
[[8]]
NULL
Pre-Processing
Now we are going to do some basic pre-processing.
data_padded_tbl <- base_data %>%
pad_by_time(
.date_var = date,
.pad_value = 0
)
# Get log interval and standardization parameters
log_params <- liv(data_padded_tbl$value, limit_lower = 0, offset = 1, silent = TRUE)
limit_lower <- log_params$limit_lower
limit_upper <- log_params$limit_upper
offset <- log_params$offset
data_liv_tbl <- data_padded_tbl %>%
# Get log interval transform
mutate(value_trans = liv(value, limit_lower = 0, offset = 1, silent = TRUE)$log_scaled)
# Get Standardization Params
std_params <- standard_vec(data_liv_tbl$value_trans, silent = TRUE)
std_mean <- std_params$mean
std_sd <- std_params$sd
data_transformed_tbl <- data_liv_tbl %>%
group_by(package) %>%
# get standardization
mutate(value_trans = standard_vec(value_trans, silent = TRUE)$standard_scaled) %>%
tk_augment_fourier(
.date_var = date,
.periods = c(7, 14, 30, 90, 180),
.K = 2
) %>%
tk_augment_timeseries_signature(
.date_var = date
) %>%
ungroup() %>%
select(-c(value, -year.iso))
Since this is panel data we can follow one of two different modeling strategies. We can search for a global model in the panel data or we can use nested forecasting finding the best model for each of the time series. Since we only have 5 panels, we will use nested forecasting.
To do this we will use the nest_timeseries and
split_nested_timeseries functions to create a nested tibble.
horizon <- 4*7
nested_data_tbl <- data_transformed_tbl %>%
# 0. Filter out column where package is NA
filter(!is.na(package)) %>%
# 1. Extending: We'll predict n days into the future.
extend_timeseries(
.id_var = package,
.date_var = date,
.length_future = horizon
) %>%
# 2. Nesting: We'll group by id, and create a future dataset
# that forecasts n days of extended data and
# an actual dataset that contains n*2 days
nest_timeseries(
.id_var = package,
.length_future = horizon
#.length_actual = horizon*2
) %>%
# 3. Splitting: We'll take the actual data and create splits
# for accuracy and confidence interval estimation of n das (test)
# and the rest is training data
split_nested_timeseries(
.length_test = horizon
)
nested_data_tbl
# A tibble: 8 × 4
package .actual_data .future_data .splits
<fct> <list> <list> <list>
1 healthyR.data <tibble [1,875 × 50]> <tibble [28 × 50]> <split [1847|28]>
2 healthyR <tibble [1,868 × 50]> <tibble [28 × 50]> <split [1840|28]>
3 healthyR.ts <tibble [1,804 × 50]> <tibble [28 × 50]> <split [1776|28]>
4 healthyverse <tibble [1,768 × 50]> <tibble [28 × 50]> <split [1740|28]>
5 healthyR.ai <tibble [1,610 × 50]> <tibble [28 × 50]> <split [1582|28]>
6 TidyDensity <tibble [1,461 × 50]> <tibble [28 × 50]> <split [1433|28]>
7 tidyAML <tibble [1,068 × 50]> <tibble [28 × 50]> <split [1040|28]>
8 RandomWalker <tibble [491 × 50]> <tibble [28 × 50]> <split [463|28]>
Now it is time to make some recipes and models using the modeltime workflow.
Modeltime Workflow
Recipe Object
recipe_base <- recipe(
value_trans ~ .
, data = extract_nested_test_split(nested_data_tbl)
)
recipe_base
recipe_date <- recipe(
value_trans ~ date
, data = extract_nested_test_split(nested_data_tbl)
)
Models
# Models ------------------------------------------------------------------
# Auto ARIMA --------------------------------------------------------------
model_spec_arima_no_boost <- arima_reg() %>%
set_engine(engine = "auto_arima")
wflw_auto_arima <- workflow() %>%
add_recipe(recipe = recipe_date) %>%
add_model(model_spec_arima_no_boost)
# NNETAR ------------------------------------------------------------------
model_spec_nnetar <- nnetar_reg(
mode = "regression"
, seasonal_period = "auto"
) %>%
set_engine("nnetar")
wflw_nnetar <- workflow() %>%
add_recipe(recipe = recipe_base) %>%
add_model(model_spec_nnetar)
# TSLM --------------------------------------------------------------------
model_spec_lm <- linear_reg() %>%
set_engine("lm")
wflw_lm <- workflow() %>%
add_recipe(recipe = recipe_base) %>%
add_model(model_spec_lm)
# MARS --------------------------------------------------------------------
model_spec_mars <- mars(mode = "regression") %>%
set_engine("earth")
wflw_mars <- workflow() %>%
add_recipe(recipe = recipe_date) %>%
add_model(model_spec_mars)
Nested Modeltime Tables
nested_modeltime_tbl <- modeltime_nested_fit(
# Nested Data
nested_data = nested_data_tbl,
control = control_nested_fit(
verbose = TRUE,
allow_par = FALSE
),
# Add workflows
wflw_auto_arima,
wflw_lm,
wflw_mars,
wflw_nnetar
)
nested_modeltime_tbl <- nested_modeltime_tbl[!is.na(nested_modeltime_tbl$package),]
Model Accuracy
nested_modeltime_tbl %>%
extract_nested_test_accuracy() %>%
filter(!is.na(package)) %>%
knitr::kable()
| package | .model_id | .model_desc | .type | mae | mape | mase | smape | rmse | rsq |
|---|---|---|---|---|---|---|---|---|---|
| healthyR.data | 1 | ARIMA | Test | 0.7861257 | 148.51882 | 0.7325263 | 148.32190 | 0.9765162 | 0.0347229 |
| healthyR.data | 2 | LM | Test | 0.6707149 | 122.03877 | 0.6249844 | 152.76197 | 0.8591373 | 0.0828554 |
| healthyR.data | 3 | EARTH | Test | 0.7128457 | 140.62649 | 0.6642426 | 138.87775 | 0.9087615 | 0.0201941 |
| healthyR.data | 4 | NNAR | Test | 0.7714193 | 177.33233 | 0.7188226 | 153.52162 | 0.9453784 | 0.0019735 |
| healthyR | 1 | ARIMA | Test | 0.6969470 | 444.10659 | 0.5719717 | 119.55143 | 0.9332107 | 0.0306813 |
| healthyR | 2 | LM | Test | 0.7253123 | 498.22767 | 0.5952506 | 125.46270 | 0.9470043 | 0.0775172 |
| healthyR | 3 | EARTH | Test | 0.6998876 | 616.60440 | 0.5743850 | 109.35748 | 0.9368951 | 0.0006382 |
| healthyR | 4 | NNAR | Test | 0.7558256 | 540.09648 | 0.6202923 | 128.67513 | 0.9898973 | 0.0450697 |
| healthyR.ts | 1 | ARIMA | Test | 1.0993203 | 100.60224 | 0.8326380 | 162.45130 | 1.3690996 | 0.0070127 |
| healthyR.ts | 2 | LM | Test | 1.1915530 | 144.62787 | 0.9024962 | 151.39576 | 1.4899607 | 0.0445971 |
| healthyR.ts | 3 | EARTH | Test | 1.0103668 | 275.44845 | 0.7652636 | 112.56529 | 1.2829877 | 0.1413370 |
| healthyR.ts | 4 | NNAR | Test | 1.2171139 | 171.73169 | 0.9218563 | 147.75712 | 1.5526050 | 0.1485170 |
| healthyverse | 1 | ARIMA | Test | 1.1924345 | 84.36053 | 1.4340005 | 145.61504 | 1.3771293 | 0.0050349 |
| healthyverse | 2 | LM | Test | 1.1184336 | 92.48080 | 1.3450084 | 128.48236 | 1.2851467 | 0.0731136 |
| healthyverse | 3 | EARTH | Test | 2.7519753 | 238.18463 | 3.3094765 | 191.49688 | 3.0097494 | 0.1297871 |
| healthyverse | 4 | NNAR | Test | 1.1991143 | 98.25826 | 1.4420335 | 143.35471 | 1.3670193 | 0.0806083 |
| healthyR.ai | 1 | ARIMA | Test | 0.7002754 | 87.59895 | 0.8389962 | 148.64846 | 0.8035017 | 0.1250908 |
| healthyR.ai | 2 | LM | Test | 0.7826767 | 130.76340 | 0.9377207 | 146.60331 | 0.9133871 | 0.0858783 |
| healthyR.ai | 3 | EARTH | Test | 1.7216247 | 517.51167 | 2.0626693 | 115.10396 | 1.9141738 | 0.0268126 |
| healthyR.ai | 4 | NNAR | Test | 0.7667847 | 137.70574 | 0.9186806 | 139.25446 | 0.9020992 | 0.0324612 |
| TidyDensity | 1 | ARIMA | Test | 1.0199895 | 122.09842 | 0.6090573 | 167.28801 | 1.2569990 | 0.0000543 |
| TidyDensity | 2 | LM | Test | 1.1053368 | 312.53278 | 0.6600199 | 162.83046 | 1.2041230 | 0.0459268 |
| TidyDensity | 3 | EARTH | Test | 1.0092686 | 140.58837 | 0.6026556 | 137.41413 | 1.3054063 | 0.0043465 |
| TidyDensity | 4 | NNAR | Test | 1.0629201 | 293.91149 | 0.6346920 | 160.18976 | 1.1954292 | 0.0481894 |
| tidyAML | 1 | ARIMA | Test | 0.6202493 | 108.98397 | 0.6070450 | 84.48303 | 0.8141327 | 0.1091012 |
| tidyAML | 2 | LM | Test | 1.0905811 | 167.31951 | 1.0673641 | 156.56941 | 1.3428815 | 0.0242122 |
| tidyAML | 3 | EARTH | Test | 2.7924093 | 906.03109 | 2.7329626 | 133.53045 | 3.1029415 | 0.0396336 |
| tidyAML | 4 | NNAR | Test | 0.8032429 | 169.67553 | 0.7861429 | 140.45166 | 1.0598661 | 0.0180034 |
| RandomWalker | 1 | ARIMA | Test | 1.0809094 | 120.73027 | 0.6529512 | 185.95200 | 1.1759656 | 0.0907719 |
| RandomWalker | 2 | LM | Test | 1.1807757 | 217.39646 | 0.7132779 | 168.27049 | 1.2366247 | 0.0214180 |
| RandomWalker | 3 | EARTH | Test | 0.9472123 | 103.43790 | 0.5721880 | 147.73908 | 1.1311154 | 0.0007775 |
| RandomWalker | 4 | NNAR | Test | 1.2054573 | 205.14477 | 0.7281875 | 177.03437 | 1.2612058 | 0.0378941 |
Plot Models
nested_modeltime_tbl %>%
extract_nested_test_forecast() %>%
group_by(package) %>%
filter_by_time(.date_var = .index, .start_date = max(.index) - 60) %>%
ungroup() %>%
plot_modeltime_forecast(
.interactive = FALSE,
.conf_interval_show = FALSE,
.facet_scales = "free"
) +
theme_minimal() +
facet_wrap(~ package, nrow = 3) +
theme(legend.position = "bottom")
Best Model
best_nested_modeltime_tbl <- nested_modeltime_tbl %>%
modeltime_nested_select_best(
metric = "rmse",
minimize = TRUE,
filter_test_forecasts = TRUE
)
best_nested_modeltime_tbl %>%
extract_nested_best_model_report()
# Nested Modeltime Table
# A tibble: 8 × 10
package .model_id .model_desc .type mae mape mase smape rmse rsq
<fct> <int> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 healthyR.da… 2 LM Test 0.671 122. 0.625 153. 0.859 8.29e-2
2 healthyR 1 ARIMA Test 0.697 444. 0.572 120. 0.933 3.07e-2
3 healthyR.ts 3 EARTH Test 1.01 275. 0.765 113. 1.28 1.41e-1
4 healthyverse 2 LM Test 1.12 92.5 1.35 128. 1.29 7.31e-2
5 healthyR.ai 1 ARIMA Test 0.700 87.6 0.839 149. 0.804 1.25e-1
6 TidyDensity 4 NNAR Test 1.06 294. 0.635 160. 1.20 4.82e-2
7 tidyAML 1 ARIMA Test 0.620 109. 0.607 84.5 0.814 1.09e-1
8 RandomWalker 3 EARTH Test 0.947 103. 0.572 148. 1.13 7.77e-4
best_nested_modeltime_tbl %>%
extract_nested_test_forecast() %>%
#filter(!is.na(.model_id)) %>%
group_by(package) %>%
filter_by_time(.date_var = .index, .start_date = max(.index) - 60) %>%
ungroup() %>%
plot_modeltime_forecast(
.interactive = FALSE,
.conf_interval_alpha = 0.2,
.facet_scales = "free"
) +
facet_wrap(~ package, nrow = 3) +
theme_minimal() +
theme(legend.position = "bottom")
Refitting and Future Forecast
Now that we have the best models, we can make our future forecasts.
nested_modeltime_refit_tbl <- best_nested_modeltime_tbl %>%
modeltime_nested_refit(
control = control_nested_refit(verbose = TRUE)
)
nested_modeltime_refit_tbl
# Nested Modeltime Table
# A tibble: 8 × 5
package .actual_data .future_data .splits .modeltime_tables
<fct> <list> <list> <list> <list>
1 healthyR.data <tibble> <tibble> <split [1847|28]> <mdl_tm_t [1 × 5]>
2 healthyR <tibble> <tibble> <split [1840|28]> <mdl_tm_t [1 × 5]>
3 healthyR.ts <tibble> <tibble> <split [1776|28]> <mdl_tm_t [1 × 5]>
4 healthyverse <tibble> <tibble> <split [1740|28]> <mdl_tm_t [1 × 5]>
5 healthyR.ai <tibble> <tibble> <split [1582|28]> <mdl_tm_t [1 × 5]>
6 TidyDensity <tibble> <tibble> <split [1433|28]> <mdl_tm_t [1 × 5]>
7 tidyAML <tibble> <tibble> <split [1040|28]> <mdl_tm_t [1 × 5]>
8 RandomWalker <tibble> <tibble> <split [463|28]> <mdl_tm_t [1 × 5]>
nested_modeltime_refit_tbl %>%
extract_nested_future_forecast() %>%
group_by(package) %>%
mutate(across(.value:.conf_hi, .fns = ~ standard_inv_vec(
x = .,
mean = std_mean,
sd = std_sd
)$standard_inverse_value)) %>%
mutate(across(.value:.conf_hi, .fns = ~ liiv(
x = .,
limit_lower = limit_lower,
limit_upper = limit_upper,
offset = offset
)$rescaled_v)) %>%
filter_by_time(.date_var = .index, .start_date = max(.index) - 60) %>%
ungroup() %>%
plot_modeltime_forecast(
.interactive = FALSE,
.conf_interval_alpha = 0.2,
.facet_scales = "free"
) +
facet_wrap(~ package, nrow = 3) +
theme_minimal() +
theme(legend.position = "bottom")
