pkgdown: initial commit

.Rbuildignore

@@ -5,3 +5,6 @@
 ^README\.Rmd$
 ^doc$
 ^Meta$
+^_pkgdown\.yml$
+^docs$
+^pkgdown$

NEWS.md

@@ -0,0 +1,3 @@
+# modeltime 0.0.0.9000
+
+* Working on essential package functionality

R/dials-ts_params.R

@@ -14,11 +14,11 @@
 #' period()
 #'
 #'
-#' @name arima_params
+#' @name time_series_params
 
 
 #' @export
-#' @rdname arima_params
+#' @rdname time_series_params
 period <- function(range = c(1L, 12L), trans = NULL) {
     dials::new_quant_param(
         type = "integer",

README.Rmd

@@ -21,7 +21,12 @@ knitr::opts_chunk$set(
 <!-- badges: start -->
 <!-- badges: end -->
 
-A scalable time series forecasting framework and workflow for use with the `tidymodels` ecosystem.
+A scalable time series forecasting toolset that __combines classical algorithms and machine learning into 1 framework.__
+
+```{r, echo=F, out.width='70%', fig.align='center'}
+knitr::include_graphics("vignettes/forecast_plot.jpg")
+```
+
 
 ## Features & Benefits
 
@@ -30,100 +35,11 @@ Modeltime has a few key features and benefits:
 1. __Interactive Plotting by default__ - Modeling algorithms have strengths and weaknesses that come to light when visualized interactively. 
 2. __Use Classical Algorithms and Machine Learning Algorithms together__ - No need to switch back and forth between various frameworks. `modeltime` unlocks machine learning & classical time series analysis. 
 
-## ARIMA Example
-
-Load libraries.
-
-```{r}
-library(modeltime)
-library(tidyverse)
-library(lubridate)
-library(timetk)
-library(parsnip)
-library(rsample)
-```
-
-### Step 1 - Collect data and split into training and test sets. 
-
-```{r}
-# Data
-m750 <- m4_monthly %>% filter(id == "M750")
-
-# Split Data 80/20
-splits <- initial_time_split(m750, prop = 0.9)
-```
-
-
-### Step 2 - Fit Models
-
-Fit ARIMA models to a training set. We have included some classic time series models and some new boosted versions:
-
-- __Auto ARIMA__ (see `arima_reg()`)
-- __Auto ARIMA + XGBoost__ - XGBoosted ARIMA Errors! (see `arima_boost()`)
-
-```{r, message = F}
-# Model 1: auto_arima ----
-model_fit_no_boost <- arima_reg() %>%
-    set_engine(engine = "auto_arima") %>%
-    fit(value ~ date, data = training(splits))
-
-# Model 2: arima_boost ----
-model_fit_boosted <- arima_boost(
-    min_n = 2,
-    learn_rate = 0.015
-) %>%
-    set_engine(engine = "auto_arima_xgboost") %>%
-    fit(value ~ date + as.numeric(date) + month(date, label = TRUE),
-        data = training(splits))
-```
 
-### Step 3 - Add fitted models to a Model Table. 
+## Tutorials
 
-Add as many models as you'd like. The analysis scales with multiple models. 
+- Getting Started - A walkthrough of the 6-Step Process for using `modeltime` to forecast
 
-```{r, paged.print = FALSE}
-models_tbl <- modeltime_table(
-    model_fit_no_boost,
-    model_fit_boosted
-)
-
-models_tbl
-```
-
-### Step 4 - Calibrate the model to a training set. 
-
-Calibrating adds a new column, `.calibration_data`, with the test predictions and residuals inside. This is how confidence intervals and accuracy metrics are determined from out-of-sample predictions.
-
-```{r, paged.print = FALSE}
-calibration_tbl <- models_tbl %>%
-    modeltime_calibrate(new_data = testing(splits))
-
-calibration_tbl
-```
-
-### Step 5 - Forecast vs Testing Set
-
-Visualizing the Test Error is easy to do. And you get an out-of-sample confidence interval. 
-
-```{r}
-calibration_tbl %>%
-    modeltime_forecast(
-        new_data    = testing(splits),
-        actual_data = m750
-    ) %>%
-    plot_modeltime_forecast(.interactive = FALSE)
-```
-
-### Step 6 - Refit to Full Dataset & Forecast Forward
-
-```{r, message=F, paged.print = FALSE}
-refit_tbl <- calibration_tbl %>%
-    modeltime_refit(data = m750)
-
-refit_tbl %>%
-    modeltime_forecast(h = "3 years", actual_data = m750) %>%
-    plot_modeltime_forecast(.interactive = FALSE)
-```
 
 
 

README.md

@@ -7,8 +7,10 @@
 
 <!-- badges: end -->
 
-A scalable time series forecasting framework and workflow for use with
-the `tidymodels` ecosystem.
+A scalable time series forecasting toolset that **combines classical
+algorithms and machine learning into 1 framework.**
+
+<img src="vignettes/forecast_plot.jpg" width="70%" style="display: block; margin: auto;" />
 
 ## Features & Benefits
 
@@ -22,123 +24,10 @@ Modeltime has a few key features and benefits:
     frameworks. `modeltime` unlocks machine learning & classical time
     series analysis.
 
-## ARIMA Example
-
-Load libraries.
-
-``` r
-library(modeltime)
-library(tidyverse)
-library(lubridate)
-library(timetk)
-library(parsnip)
-library(rsample)
-```
-
-### Step 1 - Collect data and split into training and test sets.
-
-``` r
-# Data
-m750 <- m4_monthly %>% filter(id == "M750")
-
-# Split Data 80/20
-splits <- initial_time_split(m750, prop = 0.9)
-```
-
-### Step 2 - Fit Models
-
-Fit ARIMA models to a training set. We have included some classic time
-series models and some new boosted versions:
-
-  - **Auto ARIMA** (see `arima_reg()`)
-  - **Auto ARIMA + XGBoost** - XGBoosted ARIMA Errors\! (see
-    `arima_boost()`)
-
-<!-- end list -->
-
-``` r
-# Model 1: auto_arima ----
-model_fit_no_boost <- arima_reg() %>%
-    set_engine(engine = "auto_arima") %>%
-    fit(value ~ date, data = training(splits))
-
-# Model 2: arima_boost ----
-model_fit_boosted <- arima_boost(
-    min_n = 2,
-    learn_rate = 0.015
-) %>%
-    set_engine(engine = "auto_arima_xgboost") %>%
-    fit(value ~ date + as.numeric(date) + month(date, label = TRUE),
-        data = training(splits))
-```
-
-### Step 3 - Add fitted models to a Model Table.
-
-Add as many models as you’d like. The analysis scales with multiple
-models.
-
-``` r
-models_tbl <- modeltime_table(
-    model_fit_no_boost,
-    model_fit_boosted
-)
-
-models_tbl
-#> # Modeltime Table
-#> # A tibble: 2 x 3
-#>   .model_id .model   .model_desc                              
-#>       <int> <list>   <chr>                                    
-#> 1         1 <fit[+]> ARIMA(0,1,1)(0,1,1)[12]                  
-#> 2         2 <fit[+]> ARIMA(0,1,1)(0,1,1)[12] W/ XGBOOST ERRORS
-```
-
-### Step 4 - Calibrate the model to a training set.
-
-Calibrating adds a new column, `.calibration_data`, with the test
-predictions and residuals inside. This is how confidence intervals and
-accuracy metrics are determined from out-of-sample predictions.
-
-``` r
-calibration_tbl <- models_tbl %>%
-    modeltime_calibrate(new_data = testing(splits))
-
-calibration_tbl
-#> # Modeltime Table
-#> # A tibble: 2 x 5
-#>   .model_id .model   .model_desc                          .type .calibration_da…
-#>       <int> <list>   <chr>                                <chr> <list>          
-#> 1         1 <fit[+]> ARIMA(0,1,1)(0,1,1)[12]              Test  <tibble [31 × 4…
-#> 2         2 <fit[+]> ARIMA(0,1,1)(0,1,1)[12] W/ XGBOOST … Test  <tibble [31 × 4…
-```
-
-### Step 5 - Forecast vs Testing Set
-
-Visualizing the Test Error is easy to do. And you get an out-of-sample
-confidence interval.
-
-``` r
-calibration_tbl %>%
-    modeltime_forecast(
-        new_data    = testing(splits),
-        actual_data = m750
-    ) %>%
-    plot_modeltime_forecast(.interactive = FALSE)
-```
-
-<img src="man/figures/README-unnamed-chunk-7-1.png" width="100%" />
-
-### Step 6 - Refit to Full Dataset & Forecast Forward
-
-``` r
-refit_tbl <- calibration_tbl %>%
-    modeltime_refit(data = m750)
-
-refit_tbl %>%
-    modeltime_forecast(h = "3 years", actual_data = m750) %>%
-    plot_modeltime_forecast(.interactive = FALSE)
-```
+## Tutorials
 
-<img src="man/figures/README-unnamed-chunk-8-1.png" width="100%" />
+  - Getting Started - A walkthrough of the 6-Step Process for using
+    `modeltime` to forecast
 
 ## Installation
 

_pkgdown.yml

@@ -0,0 +1,65 @@
+destination: docs
+
+template:
+  params:
+    bootswatch: flatly
+    ganalytics: UA-76139189-2
+
+navbar:
+  title: "modeltime"
+  left:
+    - icon: fa-home
+      href: index.html
+    - text: "Getting Started"
+      href: articles/getting-started-with-modeltime.html
+    - text: "Key Functions"
+      href: reference/index.html
+    - text: "Articles"
+      href: articles/index.html
+      menu:
+        - text: 'Getting Started with Modeltime'
+          href: articles/getting-started-with-modeltime.html
+    - text: "News"
+      href: news/index.html
+
+  right:
+    - icon: fa-github
+      href: https://github.com/business-science/modeltime
+
+reference:
+- title: Modeltime Workflow
+  desc: The main workflow functions for scalable time series modeling.
+  contents:
+  - modeltime_table
+  - modeltime_calibrate
+  - modeltime_forecast
+  - starts_with("modeltime_")
+  - starts_with("plot_modeltime")
+  - default_forecast_accuracy_metric_set
+- title: Algorithms
+  desc: The `parsnip`-adjacent algorithms that implement time series models.
+  contents:
+  - arima_reg
+  - arima_boost
+  - exp_smoothing
+- title: Parameters
+  desc: The `dials` parameter functions that support hyperparameter tuning with `tune`.
+  contents:
+  - period
+  - starts_with("non_seasonal")
+  - error
+- title: Developer Tools
+  desc: Tools for extending `modeltime`.
+  contents:
+  - new_modeltime_bridge
+  - create_xreg_recipe
+  - juice_xreg_recipe
+  - parse_index_from_data
+  - get_model_description
+  - get_arima_description
+
+repo:
+  url:
+    home: https://github.com/business-science/modeltime
+    source: https://github.com/business-science/modeltime/blob/master/
+    issue: https://github.com/business-science/modeltime/issues/