healthyverse_tsa
Time Series Analysis, Modeling and Forecasting of the Healthyverse
Packages Steven P. Sanderson II, MPH - Date: 2025-11-02
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: 158,895
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 2025-10-31 22:51:50, the file was birthed on: 2022-07-02 23:58:17.511888, and at report knit time is 2.920289^{4} 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 | 158895 |
| 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 | 116429 | 0.27 | 5 | 5 | 0 | 48 | 0 |
| r_arch | 116429 | 0.27 | 3 | 7 | 0 | 5 | 0 |
| r_os | 116429 | 0.27 | 7 | 15 | 0 | 23 | 0 |
| package | 0 | 1.00 | 7 | 13 | 0 | 8 | 0 |
| version | 0 | 1.00 | 5 | 17 | 0 | 62 | 0 |
| country | 14869 | 0.91 | 2 | 2 | 0 | 165 | 0 |
Variable type: Date
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| date | 0 | 1 | 2020-11-23 | 2025-10-31 | 2023-10-16 | 1797 |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| size | 0 | 1 | 1125146.29 | 1492593.09 | 355 | 23083.5 | 309574 | 2356775 | 5677952 | ▇▁▂▁▁ |
| ip_id | 0 | 1 | 11304.81 | 21910.38 | 1 | 223.0 | 2938 | 11961 | 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 | 2025-10-31 22:51:50 | 2023-10-16 11:04:28 | 99873 |
Variable type: Timespan
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| time | 0 | 1 | 0 | 59 | 12H 6M 18S | 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
-148.23 -36.48 -11.00 26.90 816.46
Coefficients:
Estimate Std. Error
(Intercept) -2.081e+02 6.202e+01
date 1.248e-02 3.288e-03
lag(value, 1) 1.125e-01 2.340e-02
lag(value, 7) 9.023e-02 2.418e-02
lag(value, 14) 7.808e-02 2.420e-02
lag(value, 21) 6.980e-02 2.426e-02
lag(value, 28) 7.133e-02 2.420e-02
lag(value, 35) 6.226e-02 2.431e-02
lag(value, 42) 5.686e-02 2.444e-02
lag(value, 49) 7.369e-02 2.447e-02
month(date, label = TRUE).L -8.492e+00 5.103e+00
month(date, label = TRUE).Q 2.215e+00 5.025e+00
month(date, label = TRUE).C -1.584e+01 5.021e+00
month(date, label = TRUE)^4 -7.829e+00 5.041e+00
month(date, label = TRUE)^5 -9.214e+00 5.032e+00
month(date, label = TRUE)^6 -1.655e+00 5.073e+00
month(date, label = TRUE)^7 -6.672e+00 4.930e+00
month(date, label = TRUE)^8 -5.525e+00 4.865e+00
month(date, label = TRUE)^9 2.111e+00 4.840e+00
month(date, label = TRUE)^10 7.541e-01 4.849e+00
month(date, label = TRUE)^11 -4.312e+00 4.837e+00
fourier_vec(date, type = "sin", K = 1, period = 7) -1.108e+01 2.245e+00
fourier_vec(date, type = "cos", K = 1, period = 7) 7.068e+00 2.327e+00
t value Pr(>|t|)
(Intercept) -3.355 0.000811 ***
date 3.795 0.000153 ***
lag(value, 1) 4.806 1.67e-06 ***
lag(value, 7) 3.732 0.000196 ***
lag(value, 14) 3.227 0.001275 **
lag(value, 21) 2.877 0.004069 **
lag(value, 28) 2.948 0.003246 **
lag(value, 35) 2.561 0.010509 *
lag(value, 42) 2.327 0.020084 *
lag(value, 49) 3.011 0.002640 **
month(date, label = TRUE).L -1.664 0.096230 .
month(date, label = TRUE).Q 0.441 0.659400
month(date, label = TRUE).C -3.156 0.001629 **
month(date, label = TRUE)^4 -1.553 0.120531
month(date, label = TRUE)^5 -1.831 0.067265 .
month(date, label = TRUE)^6 -0.326 0.744230
month(date, label = TRUE)^7 -1.353 0.176116
month(date, label = TRUE)^8 -1.136 0.256232
month(date, label = TRUE)^9 0.436 0.662788
month(date, label = TRUE)^10 0.156 0.876443
month(date, label = TRUE)^11 -0.891 0.372825
fourier_vec(date, type = "sin", K = 1, period = 7) -4.934 8.82e-07 ***
fourier_vec(date, type = "cos", K = 1, period = 7) 3.037 0.002421 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 59.31 on 1725 degrees of freedom
(49 observations deleted due to missingness)
Multiple R-squared: 0.2365, Adjusted R-squared: 0.2268
F-statistic: 24.29 on 22 and 1725 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 = 15.5804295627958"
[1] "BEST method = 'lin' PATH MEMBER = c( 25 )"
[1] "BEST lin OBJECTIVE FUNCTION = 15.5804295627958"
[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 = 8.96341410993083"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 25 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 8.96341410993083"
[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 = 11.2143130732662"
[1] "BEST method = 'both' PATH MEMBER = c( 25 )"
[1] "BEST both OBJECTIVE FUNCTION = 11.2143130732662"
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( 1 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 590.741224979835"
[1] "BEST method = 'lin' PATH MEMBER = c( 1 )"
[1] "BEST lin OBJECTIVE FUNCTION = 590.741224979835"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 1 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 156.395721368508"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 1 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 156.395721368508"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 1 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 218.542195347468"
[1] "BEST method = 'both' PATH MEMBER = c( 1 )"
[1] "BEST both OBJECTIVE FUNCTION = 218.542195347468"
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( 1 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 498.129063591367"
[1] "BEST method = 'lin' PATH MEMBER = c( 1 )"
[1] "BEST lin OBJECTIVE FUNCTION = 498.129063591367"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 1 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 129.304167568464"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 1 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 129.304167568464"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 1 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 113.558305726733"
[1] "BEST method = 'both' PATH MEMBER = c( 1 )"
[1] "BEST both OBJECTIVE FUNCTION = 113.558305726733"
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 = 11.5124886039259"
[1] "BEST method = 'lin' PATH MEMBER = c( 11 )"
[1] "BEST lin OBJECTIVE FUNCTION = 11.5124886039259"
[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 = 10.7646114514462"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 11 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 10.7646114514462"
[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 = 7.79736991134872"
[1] "BEST method = 'both' PATH MEMBER = c( 11 )"
[1] "BEST both OBJECTIVE FUNCTION = 7.79736991134872"
Package: healthyverse
[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 = 19.1986964491719"
[1] "BEST method = 'lin' PATH MEMBER = c( 6 )"
[1] "BEST lin OBJECTIVE FUNCTION = 19.1986964491719"
[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 = 12.316936393815"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 6 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 12.316936393815"
[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 = 15.3148719434479"
[1] "BEST method = 'both' PATH MEMBER = c( 6 )"
[1] "BEST both OBJECTIVE FUNCTION = 15.3148719434479"
Package: RandomWalker
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 19 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 9.09180073278475"
[1] "BEST method = 'lin' PATH MEMBER = c( 19 )"
[1] "BEST lin OBJECTIVE FUNCTION = 9.09180073278475"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 19 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 4.74767708554997"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 19 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 4.74767708554997"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 19 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 4.89001128838205"
[1] "BEST method = 'both' PATH MEMBER = c( 19 )"
[1] "BEST both OBJECTIVE FUNCTION = 4.89001128838205"
Package: tidyAML
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 18 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 46.8517827452422"
[1] "BEST method = 'lin' PATH MEMBER = c( 18 )"
[1] "BEST lin OBJECTIVE FUNCTION = 46.8517827452422"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 18 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 26.0158208352096"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 18 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 26.0158208352096"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 18 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 37.6861374537695"
[1] "BEST method = 'both' PATH MEMBER = c( 18 )"
[1] "BEST both OBJECTIVE FUNCTION = 37.6861374537695"
Package: TidyDensity
[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 = 52.9480030817349"
[1] "BEST method = 'lin' PATH MEMBER = c( 13 )"
[1] "BEST lin OBJECTIVE FUNCTION = 52.9480030817349"
[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 = 24.5047389265562"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 13 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 24.5047389265562"
[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 = 41.8594838789343"
[1] "BEST method = 'both' PATH MEMBER = c( 13 )"
[1] "BEST both OBJECTIVE FUNCTION = 41.8594838789343"
[[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 [1,789 × 50]> <tibble [28 × 50]> <split [1761|28]>
2 healthyR <tibble [1,780 × 50]> <tibble [28 × 50]> <split [1752|28]>
3 healthyR.ts <tibble [1,726 × 50]> <tibble [28 × 50]> <split [1698|28]>
4 healthyverse <tibble [1,697 × 50]> <tibble [28 × 50]> <split [1669|28]>
5 healthyR.ai <tibble [1,522 × 50]> <tibble [28 × 50]> <split [1494|28]>
6 TidyDensity <tibble [1,373 × 50]> <tibble [28 × 50]> <split [1345|28]>
7 tidyAML <tibble [980 × 50]> <tibble [28 × 50]> <split [952|28]>
8 RandomWalker <tibble [403 × 50]> <tibble [28 × 50]> <split [375|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 | 0.6860566 | 94.49646 | 0.6583493 | 165.4495 | 0.8373105 | 0.0393588 |
| healthyR.data | 2 | LM | Test | 0.7686737 | 150.39057 | 0.7376298 | 143.2891 | 0.8983241 | 0.0109533 |
| healthyR.data | 3 | EARTH | Test | 0.7117346 | 104.52915 | 0.6829902 | 188.4506 | 0.8411133 | 0.0055827 |
| healthyR.data | 4 | NNAR | Test | 0.8393499 | 152.17171 | 0.8054516 | 161.7203 | 0.9756989 | 0.0386855 |
| healthyR | 1 | ARIMA | Test | 0.5639221 | 125.57805 | 0.6498882 | 178.6568 | 0.6775840 | 0.0388675 |
| healthyR | 2 | LM | Test | 0.6730510 | 194.22445 | 0.7756530 | 134.6001 | 0.8541241 | 0.0047002 |
| healthyR | 3 | EARTH | Test | 0.6234456 | 184.72431 | 0.7184856 | 130.2775 | 0.7932458 | 0.0588763 |
| healthyR | 4 | NNAR | Test | 0.6569660 | 212.97460 | 0.7571160 | 140.2323 | 0.8522134 | 0.0012272 |
| healthyR.ts | 1 | ARIMA | Test | 0.6104449 | 98.35892 | 0.6723201 | 167.0193 | 0.7668557 | 0.0624548 |
| healthyR.ts | 2 | LM | Test | 0.7574262 | 143.83200 | 0.8341997 | 151.1368 | 0.9644850 | 0.0243804 |
| healthyR.ts | 3 | EARTH | Test | 0.6626469 | 137.05704 | 0.7298134 | 158.2388 | 0.8272858 | 0.0282012 |
| healthyR.ts | 4 | NNAR | Test | 0.8260545 | 166.20969 | 0.9097842 | 151.3799 | 1.0560208 | 0.0597480 |
| healthyverse | 1 | ARIMA | Test | 0.5234617 | 98.06902 | 0.7711117 | 148.7501 | 0.6325840 | 0.0196648 |
| healthyverse | 2 | LM | Test | 0.6205639 | 172.85390 | 0.9141530 | 138.0963 | 0.7078485 | 0.1027186 |
| healthyverse | 3 | EARTH | Test | 0.5613246 | 135.96826 | 0.8268875 | 135.1150 | 0.6727175 | 0.0784386 |
| healthyverse | 4 | NNAR | Test | 0.5747948 | 140.93812 | 0.8467305 | 128.4128 | 0.6971829 | 0.0927693 |
| healthyR.ai | 1 | ARIMA | Test | 0.4099173 | 98.75522 | 0.7532129 | 162.2933 | 0.4972562 | 0.0496672 |
| healthyR.ai | 2 | LM | Test | 0.6263568 | 248.08636 | 1.1509152 | 159.2367 | 0.7615087 | 0.0480105 |
| healthyR.ai | 3 | EARTH | Test | 0.6131487 | 264.94641 | 1.1266456 | 133.1625 | 0.7246789 | 0.1290064 |
| healthyR.ai | 4 | NNAR | Test | 0.5998022 | 242.64702 | 1.1021217 | 160.0853 | 0.7147325 | 0.0737152 |
| TidyDensity | 1 | ARIMA | Test | 1.0806159 | 505.16562 | 0.7158534 | 117.5164 | 1.2031337 | 0.0371943 |
| TidyDensity | 2 | LM | Test | 1.1698583 | 185.87785 | 0.7749720 | 155.3346 | 1.4718949 | 0.0050785 |
| TidyDensity | 3 | EARTH | Test | 3.5318479 | 1431.16638 | 2.3396708 | 172.2272 | 4.0293758 | 0.1655473 |
| TidyDensity | 4 | NNAR | Test | 1.1386690 | 275.96302 | 0.7543107 | 142.1892 | 1.3743394 | 0.0001612 |
| tidyAML | 1 | ARIMA | Test | 0.9703106 | 114.85221 | 0.9711516 | 186.9399 | 1.3837520 | 0.0428769 |
| tidyAML | 2 | LM | Test | 1.0954919 | 172.68211 | 1.0964414 | 157.8769 | 1.5642754 | 0.0072444 |
| tidyAML | 3 | EARTH | Test | 0.9421739 | 132.84946 | 0.9429905 | 162.6788 | 1.3432733 | 0.1350014 |
| tidyAML | 4 | NNAR | Test | 1.0996885 | 157.91933 | 1.1006417 | 154.3496 | 1.5435356 | 0.0535841 |
| RandomWalker | 1 | ARIMA | Test | 0.6914439 | 103.07454 | 0.5631983 | 146.9817 | 0.8070400 | 0.1793714 |
| RandomWalker | 2 | LM | Test | 0.7231889 | 101.53182 | 0.5890554 | 140.5219 | 0.8968436 | 0.0055565 |
| RandomWalker | 3 | EARTH | Test | 0.7772962 | 143.46483 | 0.6331271 | 137.4004 | 0.8521892 | 0.0020101 |
| RandomWalker | 4 | NNAR | Test | 0.8306189 | 145.88454 | 0.6765598 | 161.6142 | 0.9502803 | 0.0203947 |
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.data 1 ARIMA Test 0.686 94.5 0.658 165. 0.837 0.0394
2 healthyR 1 ARIMA Test 0.564 126. 0.650 179. 0.678 0.0389
3 healthyR.ts 1 ARIMA Test 0.610 98.4 0.672 167. 0.767 0.0625
4 healthyverse 1 ARIMA Test 0.523 98.1 0.771 149. 0.633 0.0197
5 healthyR.ai 1 ARIMA Test 0.410 98.8 0.753 162. 0.497 0.0497
6 TidyDensity 1 ARIMA Test 1.08 505. 0.716 118. 1.20 0.0372
7 tidyAML 3 EARTH Test 0.942 133. 0.943 163. 1.34 0.135
8 RandomWalker 1 ARIMA Test 0.691 103. 0.563 147. 0.807 0.179
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 [1761|28]> <mdl_tm_t [1 × 5]>
2 healthyR <tibble> <tibble> <split [1752|28]> <mdl_tm_t [1 × 5]>
3 healthyR.ts <tibble> <tibble> <split [1698|28]> <mdl_tm_t [1 × 5]>
4 healthyverse <tibble> <tibble> <split [1669|28]> <mdl_tm_t [1 × 5]>
5 healthyR.ai <tibble> <tibble> <split [1494|28]> <mdl_tm_t [1 × 5]>
6 TidyDensity <tibble> <tibble> <split [1345|28]> <mdl_tm_t [1 × 5]>
7 tidyAML <tibble> <tibble> <split [952|28]> <mdl_tm_t [1 × 5]>
8 RandomWalker <tibble> <tibble> <split [375|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")
