healthyverse_tsa
Time Series Analysis, Modeling and Forecasting of the Healthyverse
Packages Steven P. Sanderson II, MPH - Date: 2025-12-07
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: 162,201
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 2025-12-05 23:44:53, the file was birthed on: 2022-07-02 23:58:17.511888, and at report knit time is 3.004378^{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 | 162201 |
| 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 | 118930 | 0.27 | 5 | 7 | 0 | 50 | 0 |
| r_arch | 118930 | 0.27 | 1 | 7 | 0 | 6 | 0 |
| r_os | 118930 | 0.27 | 7 | 19 | 0 | 24 | 0 |
| package | 0 | 1.00 | 7 | 13 | 0 | 8 | 0 |
| version | 0 | 1.00 | 5 | 17 | 0 | 62 | 0 |
| country | 15208 | 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 | 2025-12-05 | 2023-11-03 | 1832 |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| size | 0 | 1 | 1124686.85 | 1488502.62 | 355 | 27384 | 310267 | 2354018 | 5677952 | ▇▁▂▁▁ |
| ip_id | 0 | 1 | 11338.33 | 21980.46 | 1 | 236 | 2898 | 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 | 2025-12-05 23:44:53 | 2023-11-03 09:08:52 | 102373 |
Variable type: Timespan
| skim_variable | n_missing | complete_rate | min | max | median | n_unique |
|---|---|---|---|---|---|---|
| time | 0 | 1 | 0 | 59 | 12H 6M 46S | 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
-146.49 -36.47 -11.11 27.22 819.51
Coefficients:
Estimate Std. Error
(Intercept) -1.791e+02 6.058e+01
date 1.100e-02 3.211e-03
lag(value, 1) 1.102e-01 2.321e-02
lag(value, 7) 8.971e-02 2.391e-02
lag(value, 14) 7.643e-02 2.389e-02
lag(value, 21) 8.058e-02 2.396e-02
lag(value, 28) 6.884e-02 2.389e-02
lag(value, 35) 5.546e-02 2.394e-02
lag(value, 42) 6.197e-02 2.406e-02
lag(value, 49) 6.258e-02 2.396e-02
month(date, label = TRUE).L -1.084e+01 5.031e+00
month(date, label = TRUE).Q 1.779e-01 4.961e+00
month(date, label = TRUE).C -1.615e+01 4.989e+00
month(date, label = TRUE)^4 -6.314e+00 4.954e+00
month(date, label = TRUE)^5 -6.855e+00 4.905e+00
month(date, label = TRUE)^6 1.163e+00 4.913e+00
month(date, label = TRUE)^7 -4.552e+00 4.847e+00
month(date, label = TRUE)^8 -4.039e+00 4.819e+00
month(date, label = TRUE)^9 2.758e+00 4.832e+00
month(date, label = TRUE)^10 9.211e-01 4.849e+00
month(date, label = TRUE)^11 -4.078e+00 4.835e+00
fourier_vec(date, type = "sin", K = 1, period = 7) -1.129e+01 2.217e+00
fourier_vec(date, type = "cos", K = 1, period = 7) 7.259e+00 2.295e+00
t value Pr(>|t|)
(Intercept) -2.956 0.003157 **
date 3.426 0.000627 ***
lag(value, 1) 4.750 2.20e-06 ***
lag(value, 7) 3.751 0.000182 ***
lag(value, 14) 3.199 0.001404 **
lag(value, 21) 3.364 0.000785 ***
lag(value, 28) 2.881 0.004006 **
lag(value, 35) 2.317 0.020605 *
lag(value, 42) 2.575 0.010097 *
lag(value, 49) 2.612 0.009079 **
month(date, label = TRUE).L -2.154 0.031395 *
month(date, label = TRUE).Q 0.036 0.971397
month(date, label = TRUE).C -3.237 0.001231 **
month(date, label = TRUE)^4 -1.274 0.202679
month(date, label = TRUE)^5 -1.398 0.162404
month(date, label = TRUE)^6 0.237 0.812936
month(date, label = TRUE)^7 -0.939 0.347768
month(date, label = TRUE)^8 -0.838 0.402060
month(date, label = TRUE)^9 0.571 0.568212
month(date, label = TRUE)^10 0.190 0.849351
month(date, label = TRUE)^11 -0.844 0.399036
fourier_vec(date, type = "sin", K = 1, period = 7) -5.091 3.95e-07 ***
fourier_vec(date, type = "cos", K = 1, period = 7) 3.163 0.001586 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 59.31 on 1760 degrees of freedom
(49 observations deleted due to missingness)
Multiple R-squared: 0.23, Adjusted R-squared: 0.2204
F-statistic: 23.9 on 22 and 1760 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( 12 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 6.65202338339804"
[1] "BEST method = 'lin' PATH MEMBER = c( 12 )"
[1] "BEST lin OBJECTIVE FUNCTION = 6.65202338339804"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 12 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 7.4227499485934"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 12 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 7.4227499485934"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 12 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 6.33030631089571"
[1] "BEST method = 'both' PATH MEMBER = c( 12 )"
[1] "BEST both OBJECTIVE FUNCTION = 6.33030631089571"
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 = 19.6427318071377"
[1] "BEST method = 'lin' PATH MEMBER = c( 3 )"
[1] "BEST lin OBJECTIVE FUNCTION = 19.6427318071377"
[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 = 4.20959069058834"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 3 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 4.20959069058834"
[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 = 9.20597684497577"
[1] "BEST method = 'both' PATH MEMBER = c( 3 )"
[1] "BEST both OBJECTIVE FUNCTION = 9.20597684497577"
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( 5 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 33.6693503968154"
[1] "BEST method = 'lin' PATH MEMBER = c( 5 )"
[1] "BEST lin OBJECTIVE FUNCTION = 33.6693503968154"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 5 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 11.2897038891863"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 5 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 11.2897038891863"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 5 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 12.6461274907765"
[1] "BEST method = 'both' PATH MEMBER = c( 5 )"
[1] "BEST both OBJECTIVE FUNCTION = 12.6461274907765"
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( 1 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 111.487808874278"
[1] "BEST method = 'lin' PATH MEMBER = c( 1 )"
[1] "BEST lin OBJECTIVE FUNCTION = 111.487808874278"
[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 = 99.9041722125259"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 1 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 99.9041722125259"
[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 = 261.956598172143"
[1] "BEST method = 'both' PATH MEMBER = c( 1 )"
[1] "BEST both OBJECTIVE FUNCTION = 261.956598172143"
Package: healthyverse
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 12 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 5.83528712399909"
[1] "BEST method = 'lin' PATH MEMBER = c( 12 )"
[1] "BEST lin OBJECTIVE FUNCTION = 5.83528712399909"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 12 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 9.92096738719668"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 12 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 9.92096738719668"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 12 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 8.14933892380053"
[1] "BEST method = 'both' PATH MEMBER = c( 12 )"
[1] "BEST both OBJECTIVE FUNCTION = 8.14933892380053"
Package: RandomWalker
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 14 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 8.16119090785943"
[1] "BEST method = 'lin' PATH MEMBER = c( 14 )"
[1] "BEST lin OBJECTIVE FUNCTION = 8.16119090785943"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 14 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 7.68876117911321"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 14 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 7.68876117911321"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 14 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 7.08249875413898"
[1] "BEST method = 'both' PATH MEMBER = c( 14 )"
[1] "BEST both OBJECTIVE FUNCTION = 7.08249875413898"
Package: tidyAML
[1] "CURRNET METHOD: lin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'lin' , seasonal.factor = c( 5 ) ...)"
[1] "CURRENT lin OBJECTIVE FUNCTION = 18.353612865021"
[1] "BEST method = 'lin' PATH MEMBER = c( 5 )"
[1] "BEST lin OBJECTIVE FUNCTION = 18.353612865021"
[1] "CURRNET METHOD: nonlin"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'nonlin' , seasonal.factor = c( 5 ) ...)"
[1] "CURRENT nonlin OBJECTIVE FUNCTION = 20.8297652176163"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 5 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 20.8297652176163"
[1] "CURRNET METHOD: both"
[1] "COPY LATEST PARAMETERS DIRECTLY FOR NNS.ARMA() IF ERROR:"
[1] "NNS.ARMA(... method = 'both' , seasonal.factor = c( 5 ) ...)"
[1] "CURRENT both OBJECTIVE FUNCTION = 17.9414486545205"
[1] "BEST method = 'both' PATH MEMBER = c( 5 )"
[1] "BEST both OBJECTIVE FUNCTION = 17.9414486545205"
Package: TidyDensity
[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 = 8.38071347350636"
[1] "BEST method = 'lin' PATH MEMBER = c( 3 )"
[1] "BEST lin OBJECTIVE FUNCTION = 8.38071347350636"
[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 = 254.843564687428"
[1] "BEST method = 'nonlin' PATH MEMBER = c( 3 )"
[1] "BEST nonlin OBJECTIVE FUNCTION = 254.843564687428"
[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 = 161.842384677563"
[1] "BEST method = 'both' PATH MEMBER = c( 3 )"
[1] "BEST both OBJECTIVE FUNCTION = 161.842384677563"
[[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 [1,824 × 50]> <tibble [28 × 50]> <split [1796|28]>
2 healthyR <tibble [1,815 × 50]> <tibble [28 × 50]> <split [1787|28]>
3 healthyR.ts <tibble [1,760 × 50]> <tibble [28 × 50]> <split [1732|28]>
4 healthyverse <tibble [1,731 × 50]> <tibble [28 × 50]> <split [1703|28]>
5 healthyR.ai <tibble [1,557 × 50]> <tibble [28 × 50]> <split [1529|28]>
6 TidyDensity <tibble [1,408 × 50]> <tibble [28 × 50]> <split [1380|28]>
7 tidyAML <tibble [1,015 × 50]> <tibble [28 × 50]> <split [987|28]>
8 RandomWalker <tibble [438 × 50]> <tibble [28 × 50]> <split [410|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 | 0.8437691 | 99.42201 | 0.7287151 | 157.1201 | 1.0198470 | 0.0469323 |
| healthyR.data | 2 | LM | Test | 0.7931584 | 175.99518 | 0.6850055 | 127.9006 | 0.9222167 | 0.0197151 |
| healthyR.data | 3 | EARTH | Test | 0.8744510 | 109.25415 | 0.7552133 | 170.7399 | 1.0406576 | 0.0031915 |
| healthyR.data | 4 | NNAR | Test | 0.7850806 | 189.69710 | 0.6780292 | 116.4043 | 0.9099128 | 0.0172959 |
| healthyR | 1 | ARIMA | Test | 0.7766232 | 179.49445 | 0.7947763 | 145.4135 | 1.0164268 | 0.0011110 |
| healthyR | 2 | LM | Test | 0.7104717 | 460.85924 | 0.7270786 | 119.9175 | 0.8780073 | 0.0449694 |
| healthyR | 3 | EARTH | Test | 0.6499397 | 488.49336 | 0.6651316 | 101.1310 | 0.8482091 | 0.0009516 |
| healthyR | 4 | NNAR | Test | 0.7674614 | 389.23537 | 0.7854003 | 132.6854 | 0.9280453 | 0.0191197 |
| healthyR.ts | 1 | ARIMA | Test | 0.7524843 | 164.02082 | 0.7565155 | 148.6750 | 0.9674465 | 0.0509100 |
| healthyR.ts | 2 | LM | Test | 0.8899913 | 269.12436 | 0.8947591 | 160.0044 | 1.0631022 | 0.0001180 |
| healthyR.ts | 3 | EARTH | Test | 0.7113556 | 115.80316 | 0.7151665 | 141.7486 | 0.9315377 | 0.0046804 |
| healthyR.ts | 4 | NNAR | Test | 0.9731767 | 320.28453 | 0.9783901 | 167.8926 | 1.1355483 | 0.0018048 |
| healthyverse | 1 | ARIMA | Test | 0.8851245 | 100.51380 | 0.9601825 | 152.2649 | 1.1145885 | 0.0027465 |
| healthyverse | 2 | LM | Test | 0.9193487 | 148.10379 | 0.9973089 | 143.7200 | 1.0626309 | 0.0016053 |
| healthyverse | 3 | EARTH | Test | 0.8245781 | 185.23231 | 0.8945018 | 111.3741 | 0.9857874 | 0.0396456 |
| healthyverse | 4 | NNAR | Test | 0.8729017 | 143.32096 | 0.9469233 | 136.6571 | 1.0453630 | 0.0086533 |
| healthyR.ai | 1 | ARIMA | Test | 0.9752663 | 103.79897 | 1.1004585 | 185.4406 | 1.1536264 | 0.0117207 |
| healthyR.ai | 2 | LM | Test | 1.1954929 | 155.20581 | 1.3489550 | 164.0372 | 1.3847690 | 0.0849085 |
| healthyR.ai | 3 | EARTH | Test | 1.4913894 | 191.98985 | 1.6828350 | 178.3417 | 1.6878221 | 0.0348017 |
| healthyR.ai | 4 | NNAR | Test | 1.1979156 | 167.12845 | 1.3516887 | 159.7321 | 1.3736177 | 0.1092835 |
| TidyDensity | 1 | ARIMA | Test | 0.9746038 | 242.87852 | 0.6462267 | 149.1447 | 1.1596522 | 0.4285271 |
| TidyDensity | 2 | LM | Test | 0.8646642 | 115.59769 | 0.5733295 | 153.3752 | 1.0505231 | 0.0471042 |
| TidyDensity | 3 | EARTH | Test | 1.2692156 | 362.01250 | 0.8415738 | 154.5490 | 1.4460913 | 0.0479124 |
| TidyDensity | 4 | NNAR | Test | 0.9810813 | 153.47045 | 0.6505218 | 153.0236 | 1.1549409 | 0.0102127 |
| tidyAML | 1 | ARIMA | Test | 0.7913115 | 113.31335 | 0.9395383 | 181.2607 | 0.9424719 | 0.0730200 |
| tidyAML | 2 | LM | Test | 0.6986177 | 215.88895 | 0.8294814 | 127.6415 | 0.8472981 | 0.0584187 |
| tidyAML | 3 | EARTH | Test | 0.9273326 | 164.73050 | 1.1010387 | 183.3334 | 1.0879849 | 0.0459931 |
| tidyAML | 4 | NNAR | Test | 0.6030920 | 158.75472 | 0.7160619 | 113.8726 | 0.7818843 | 0.1073449 |
| RandomWalker | 1 | ARIMA | Test | 0.7943307 | 132.89289 | 0.7536269 | 169.1034 | 0.8673549 | 0.2114675 |
| RandomWalker | 2 | LM | Test | 0.8792741 | 151.14417 | 0.8342176 | 161.0239 | 0.9889574 | 0.0001878 |
| RandomWalker | 3 | EARTH | Test | 0.9865742 | 176.60287 | 0.9360194 | 165.9071 | 1.0411936 | 0.0714650 |
| RandomWalker | 4 | NNAR | Test | 0.9423388 | 168.46947 | 0.8940507 | 156.8178 | 1.0692042 | 0.0081323 |
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… 4 NNAR Test 0.785 190. 0.678 116. 0.910 1.73e-2
2 healthyR 3 EARTH Test 0.650 488. 0.665 101. 0.848 9.52e-4
3 healthyR.ts 3 EARTH Test 0.711 116. 0.715 142. 0.932 4.68e-3
4 healthyverse 3 EARTH Test 0.825 185. 0.895 111. 0.986 3.96e-2
5 healthyR.ai 1 ARIMA Test 0.975 104. 1.10 185. 1.15 1.17e-2
6 TidyDensity 2 LM Test 0.865 116. 0.573 153. 1.05 4.71e-2
7 tidyAML 4 NNAR Test 0.603 159. 0.716 114. 0.782 1.07e-1
8 RandomWalker 1 ARIMA Test 0.794 133. 0.754 169. 0.867 2.11e-1
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 [1796|28]> <mdl_tm_t [1 × 5]>
2 healthyR <tibble> <tibble> <split [1787|28]> <mdl_tm_t [1 × 5]>
3 healthyR.ts <tibble> <tibble> <split [1732|28]> <mdl_tm_t [1 × 5]>
4 healthyverse <tibble> <tibble> <split [1703|28]> <mdl_tm_t [1 × 5]>
5 healthyR.ai <tibble> <tibble> <split [1529|28]> <mdl_tm_t [1 × 5]>
6 TidyDensity <tibble> <tibble> <split [1380|28]> <mdl_tm_t [1 × 5]>
7 tidyAML <tibble> <tibble> <split [987|28]> <mdl_tm_t [1 × 5]>
8 RandomWalker <tibble> <tibble> <split [410|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")
