healthyverse_tsa
Time Series Analysis, Modeling and Forecasting of the Healthyverse
Packages Steven P. Sanderson II, MPH - Date: 2026-03-02
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: 170,572
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 2026-02-28 22:42:03, the file was birthed on: 2025-10-31 10:47:59.603742, and at report knit time is 2887.9 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 | 170572 |
| 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 | 126185 | 0.26 | 5 | 7 | 0 | 50 | 0 |
| r_arch | 126185 | 0.26 | 1 | 7 | 0 | 6 | 0 |
| r_os | 126185 | 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 | 15979 | 0.91 | 2 | 2 | 0 | 167 | 0 |
Variable type: Date
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| date | 0 | 1 | 2020-11-23 | 2026-02-28 | 2023-12-18 | 1917 |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| size | 0 | 1 | 1127626.46 | 1482838.9 | 355 | 42213 | 323098 | 2348372 | 5677952 | ▇▁▂▁▁ |
| ip_id | 0 | 1 | 11466.42 | 22865.6 | 1 | 198 | 2759 | 11772 | 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 | 2026-02-28 22:42:03 | 2023-12-18 11:34:00 | 108246 |
Variable type: Timespan
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| time | 0 | 1 | 0 | 59 | 47 | 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
-148.77 -37.30 -11.45 27.52 826.18
Coefficients:
Estimate Std. Error
(Intercept) -1.741e+02 5.571e+01
date 1.080e-02 2.948e-03
lag(value, 1) 9.793e-02 2.276e-02
lag(value, 7) 8.529e-02 2.350e-02
lag(value, 14) 6.785e-02 2.337e-02
lag(value, 21) 8.713e-02 2.345e-02
lag(value, 28) 7.213e-02 2.339e-02
lag(value, 35) 4.648e-02 2.343e-02
lag(value, 42) 6.860e-02 2.360e-02
lag(value, 49) 6.785e-02 2.354e-02
month(date, label = TRUE).L -9.331e+00 4.759e+00
month(date, label = TRUE).Q -1.001e+00 4.737e+00
month(date, label = TRUE).C -1.436e+01 4.795e+00
month(date, label = TRUE)^4 -8.160e+00 4.806e+00
month(date, label = TRUE)^5 -5.129e+00 4.761e+00
month(date, label = TRUE)^6 -1.679e-01 4.806e+00
month(date, label = TRUE)^7 -3.649e+00 4.790e+00
month(date, label = TRUE)^8 -4.725e+00 4.802e+00
month(date, label = TRUE)^9 3.077e+00 4.835e+00
month(date, label = TRUE)^10 1.017e+00 4.856e+00
month(date, label = TRUE)^11 -4.225e+00 4.843e+00
fourier_vec(date, type = "sin", K = 1, period = 7) -1.103e+01 2.153e+00
fourier_vec(date, type = "cos", K = 1, period = 7) 7.517e+00 2.226e+00
t value Pr(>|t|)
(Intercept) -3.125 0.001808 **
date 3.663 0.000256 ***
lag(value, 1) 4.303 1.77e-05 ***
lag(value, 7) 3.630 0.000291 ***
lag(value, 14) 2.903 0.003743 **
lag(value, 21) 3.716 0.000208 ***
lag(value, 28) 3.083 0.002076 **
lag(value, 35) 1.984 0.047402 *
lag(value, 42) 2.906 0.003702 **
lag(value, 49) 2.882 0.003994 **
month(date, label = TRUE).L -1.961 0.050045 .
month(date, label = TRUE).Q -0.211 0.832661
month(date, label = TRUE).C -2.995 0.002783 **
month(date, label = TRUE)^4 -1.698 0.089674 .
month(date, label = TRUE)^5 -1.077 0.281475
month(date, label = TRUE)^6 -0.035 0.972125
month(date, label = TRUE)^7 -0.762 0.446325
month(date, label = TRUE)^8 -0.984 0.325267
month(date, label = TRUE)^9 0.636 0.524595
month(date, label = TRUE)^10 0.209 0.834184
month(date, label = TRUE)^11 -0.872 0.383169
fourier_vec(date, type = "sin", K = 1, period = 7) -5.121 3.36e-07 ***
fourier_vec(date, type = "cos", K = 1, period = 7) 3.376 0.000750 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 59.42 on 1845 degrees of freedom
(49 observations deleted due to missingness)
Multiple R-squared: 0.2199, Adjusted R-squared: 0.2106
F-statistic: 23.64 on 22 and 1845 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( 7 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 6.37356608194594"
[1] "BEST method = 'lin' PATH MEMBER = c( 7 )"
[1] "BEST lin OBJECTIVE FUNCTION = 6.37356608194594"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 7 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 11.276144093962"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 7 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 11.276144093962"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 7 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 10.4470270380261"
[1] "BEST method = 'both' PATH MEMBER = c( 7 )"
[1] "BEST both OBJECTIVE FUNCTION = 10.4470270380261"
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( 15 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 15.5761345031535"
[1] "BEST method = 'lin' PATH MEMBER = c( 15 )"
[1] "BEST lin OBJECTIVE FUNCTION = 15.5761345031535"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 15 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 13.6925188928243"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 15 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 13.6925188928243"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 15 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 15.5691403591329"
[1] "BEST method = 'both' PATH MEMBER = c( 15 )"
[1] "BEST both OBJECTIVE FUNCTION = 15.5691403591329"
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( 25 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 11.2297407466459"
[1] "BEST method = 'lin' PATH MEMBER = c( 25 )"
[1] "BEST lin OBJECTIVE FUNCTION = 11.2297407466459"
[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 = 7.35216040993209"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 25 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 7.35216040993209"
[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 = 8.5960448710708"
[1] "BEST method = 'both' PATH MEMBER = c( 25 )"
[1] "BEST both OBJECTIVE FUNCTION = 8.5960448710708"
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( 13 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 12.2490300944924"
[1] "BEST method = 'lin' PATH MEMBER = c( 13 )"
[1] "BEST lin OBJECTIVE FUNCTION = 12.2490300944924"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 13 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 8.4972392489969"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 13 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 8.4972392489969"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 13 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 11.1541695160947"
[1] "BEST method = 'both' PATH MEMBER = c( 13 )"
[1] "BEST both OBJECTIVE FUNCTION = 11.1541695160947"
Package: healthyverse
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 17 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 7.6773721804901"
[1] "BEST method = 'lin' PATH MEMBER = c( 17 )"
[1] "BEST lin OBJECTIVE FUNCTION = 7.6773721804901"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 17 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 8.9302600030279"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 17 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 8.9302600030279"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 17 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 9.01954473383587"
[1] "BEST method = 'both' PATH MEMBER = c( 17 )"
[1] "BEST both OBJECTIVE FUNCTION = 9.01954473383587"
Package: RandomWalker
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 15 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 6.85484439393979"
[1] "BEST method = 'lin' PATH MEMBER = c( 15 )"
[1] "BEST lin OBJECTIVE FUNCTION = 6.85484439393979"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 15 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 1.78894550762179"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 15 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 1.78894550762179"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 15 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 2.28804103610257"
[1] "BEST method = 'both' PATH MEMBER = c( 15 )"
[1] "BEST both OBJECTIVE FUNCTION = 2.28804103610257"
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 = 9.82983930807719"
[1] "BEST method = 'lin' PATH MEMBER = c( 8 )"
[1] "BEST lin OBJECTIVE FUNCTION = 9.82983930807719"
[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 = 21.4039946090901"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 8 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 21.4039946090901"
[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 = 14.6132765115356"
[1] "BEST method = 'both' PATH MEMBER = c( 8 )"
[1] "BEST both OBJECTIVE FUNCTION = 14.6132765115356"
Package: TidyDensity
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 6 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 9.01649478250118"
[1] "BEST method = 'lin' PATH MEMBER = c( 6 )"
[1] "BEST lin OBJECTIVE FUNCTION = 9.01649478250118"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 6 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 3.91739498306544"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 6 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 3.91739498306544"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 6 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 5.06940134663112"
[1] "BEST method = 'both' PATH MEMBER = c( 6 )"
[1] "BEST both OBJECTIVE FUNCTION = 5.06940134663112"
[[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 [1,907 × 50]> <tibble [28 × 50]> <split [1879|28]>
2 healthyR <tibble [1,900 × 50]> <tibble [28 × 50]> <split [1872|28]>
3 healthyR.ts <tibble [1,836 × 50]> <tibble [28 × 50]> <split [1808|28]>
4 healthyverse <tibble [1,791 × 50]> <tibble [28 × 50]> <split [1763|28]>
5 healthyR.ai <tibble [1,642 × 50]> <tibble [28 × 50]> <split [1614|28]>
6 TidyDensity <tibble [1,493 × 50]> <tibble [28 × 50]> <split [1465|28]>
7 tidyAML <tibble [1,099 × 50]> <tibble [28 × 50]> <split [1071|28]>
8 RandomWalker <tibble [523 × 50]> <tibble [28 × 50]> <split [495|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 | 0.7464457 | 126.83384 | 0.8118744 | 176.16792 | 0.8597226 | 0.0062263 |
| healthyR.data | 2 | LM | Test | 0.7738882 | 144.84529 | 0.8417223 | 159.24295 | 0.8970849 | 0.0456018 |
| healthyR.data | 3 | EARTH | Test | 0.8042315 | 144.28416 | 0.8747254 | 168.74712 | 0.9136580 | 0.0082615 |
| healthyR.data | 4 | NNAR | Test | 0.8002208 | 151.50062 | 0.8703631 | 156.63128 | 0.9188907 | 0.0324949 |
| healthyR | 1 | ARIMA | Test | 0.7391665 | 514.78484 | 0.7340433 | 147.30360 | 0.9018639 | 0.0450721 |
| healthyR | 2 | LM | Test | 0.6926233 | 588.57388 | 0.6878227 | 138.20546 | 0.8667681 | 0.0859462 |
| healthyR | 3 | EARTH | Test | 0.7606081 | 497.80087 | 0.7553363 | 144.85428 | 0.9308198 | 0.0173372 |
| healthyR | 4 | NNAR | Test | 0.6892321 | 444.22245 | 0.6844550 | 135.84835 | 0.8524041 | 0.1110417 |
| healthyR.ts | 1 | ARIMA | Test | 0.7075998 | 163.01846 | 0.6118779 | 159.24372 | 0.9318861 | 0.0553014 |
| healthyR.ts | 2 | LM | Test | 0.7058630 | 129.00960 | 0.6103761 | 146.61655 | 0.9641982 | 0.0366540 |
| healthyR.ts | 3 | EARTH | Test | 1.9020631 | 1238.96329 | 1.6447580 | 147.79698 | 2.2290678 | 0.0020431 |
| healthyR.ts | 4 | NNAR | Test | 0.7556262 | 214.98267 | 0.6534074 | 148.33701 | 1.0055381 | 0.0204569 |
| healthyverse | 1 | ARIMA | Test | 0.9835385 | 103.00542 | 1.2406634 | 96.95159 | 1.1261561 | 0.0227958 |
| healthyverse | 2 | LM | Test | 1.0332069 | 153.19972 | 1.3033165 | 112.71273 | 1.1472893 | 0.1260281 |
| healthyverse | 3 | EARTH | Test | 0.7132761 | 232.25465 | 0.8997467 | 57.71983 | 0.8960681 | 0.2016009 |
| healthyverse | 4 | NNAR | Test | 1.1257794 | 162.67284 | 1.4200901 | 125.96312 | 1.2757376 | 0.0354386 |
| healthyR.ai | 1 | ARIMA | Test | 0.6585147 | 164.66405 | 0.7538030 | 131.21949 | 0.8522950 | 0.0149852 |
| healthyR.ai | 2 | LM | Test | 0.7138754 | 210.12491 | 0.8171745 | 154.45338 | 0.8912640 | 0.0854100 |
| healthyR.ai | 3 | EARTH | Test | 0.7309128 | 164.65715 | 0.8366772 | 158.92415 | 0.9471818 | 0.0265064 |
| healthyR.ai | 4 | NNAR | Test | 0.7168110 | 241.12514 | 0.8205349 | 147.96642 | 0.8988304 | 0.0716873 |
| TidyDensity | 1 | ARIMA | Test | 1.1926282 | 121.59104 | 0.7518659 | 171.73820 | 1.2954543 | 0.0390283 |
| TidyDensity | 2 | LM | Test | 1.0795443 | 133.30730 | 0.6805746 | 145.69753 | 1.2146105 | 0.0800296 |
| TidyDensity | 3 | EARTH | Test | 1.1685448 | 116.10067 | 0.7366830 | 173.25895 | 1.2723721 | 0.0366156 |
| TidyDensity | 4 | NNAR | Test | 1.0869431 | 111.65039 | 0.6852390 | 150.18878 | 1.2227333 | 0.0560070 |
| tidyAML | 1 | ARIMA | Test | 0.7173749 | 209.10403 | 0.8631372 | 132.71737 | 0.8855874 | 0.0050701 |
| tidyAML | 2 | LM | Test | 0.7371603 | 208.43836 | 0.8869428 | 150.82780 | 0.8933016 | 0.0115613 |
| tidyAML | 3 | EARTH | Test | 0.7672693 | 229.83445 | 0.9231696 | 145.04370 | 0.9775633 | 0.1391832 |
| tidyAML | 4 | NNAR | Test | 0.7984529 | 205.54039 | 0.9606892 | 152.48013 | 1.0152457 | 0.0487620 |
| RandomWalker | 1 | ARIMA | Test | 0.7107075 | 96.99181 | 0.4574017 | 141.95665 | 0.8422758 | 0.4962586 |
| RandomWalker | 2 | LM | Test | 0.9038988 | 119.80396 | 0.5817371 | 133.36005 | 1.1718189 | 0.0000148 |
| RandomWalker | 3 | EARTH | Test | 0.8450776 | 89.18653 | 0.5438805 | 146.26421 | 1.0626170 | 0.0097714 |
| RandomWalker | 4 | NNAR | Test | 1.0315002 | 152.86065 | 0.6638596 | 170.51097 | 1.1837782 | 0.0731797 |
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… 1 ARIMA Test 0.746 127. 0.812 176. 0.860 0.00623
2 healthyR 4 NNAR Test 0.689 444. 0.684 136. 0.852 0.111
3 healthyR.ts 1 ARIMA Test 0.708 163. 0.612 159. 0.932 0.0553
4 healthyverse 3 EARTH Test 0.713 232. 0.900 57.7 0.896 0.202
5 healthyR.ai 1 ARIMA Test 0.659 165. 0.754 131. 0.852 0.0150
6 TidyDensity 2 LM Test 1.08 133. 0.681 146. 1.21 0.0800
7 tidyAML 1 ARIMA Test 0.717 209. 0.863 133. 0.886 0.00507
8 RandomWalker 1 ARIMA Test 0.711 97.0 0.457 142. 0.842 0.496
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 [1879|28]> <mdl_tm_t [1 × 5]>
2 healthyR <tibble> <tibble> <split [1872|28]> <mdl_tm_t [1 × 5]>
3 healthyR.ts <tibble> <tibble> <split [1808|28]> <mdl_tm_t [1 × 5]>
4 healthyverse <tibble> <tibble> <split [1763|28]> <mdl_tm_t [1 × 5]>
5 healthyR.ai <tibble> <tibble> <split [1614|28]> <mdl_tm_t [1 × 5]>
6 TidyDensity <tibble> <tibble> <split [1465|28]> <mdl_tm_t [1 × 5]>
7 tidyAML <tibble> <tibble> <split [1071|28]> <mdl_tm_t [1 × 5]>
8 RandomWalker <tibble> <tibble> <split [495|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")
