up optimize section funzionante

- error in feed-forward con decay=tune()

dashboard/R/fit_model.R

@@ -355,10 +355,10 @@ generate_model_spec <- function(method, params) {
 }
 
 # function to generate the model specification for tuning
-update_tune_model_parameters <- function(method, params) {
+set_tune_parameters <- function(method, params) {
 
-  # function to get default tuning parameters
-  get_tune <- function(parameter, value) {
+  # function to set tuning parameters
+  set_tune <- function(parameter, value) {
     if (value == FALSE) {
       get_default(parameter)
     } else {
@@ -367,19 +367,45 @@ update_tune_model_parameters <- function(method, params) {
   }
 
   mtd_params <- getOption("tsf.dashboard.methods_params")[[method]] # get the parameters for the method
-  # prm_name <- grep("_xx_", names(params), value = TRUE) # get the UI input id
-  # prm_ui_name <- eval(parse(text = paste0("params$", prm_name))) # get the UI parameters name
   if (method == "Elastic Net") {
-    prm_ui_name <- params$tune_xx_elanet
+    prm_ui_name <- params$tune_elanet
+  } else if (method == "MARS") {
+    prm_ui_name <- params$tune_mars
+  } else if (method == "KNN") {
+    prm_ui_name <- params$tune_knn
+  } else if (method == "SVM") {
+    if (params$tune_boundary == "Linear") {
+      prm_ui_name <- params$tune_svm_linear
+    } else {
+      prm_ui_name <- params$tune_svm_rbf
+    }
   } else if (method == "Random Forest") {
-    prm_ui_name <- params$tune_xx_rf
+    prm_ui_name <- params$tune_rf
+  } else if (method == "Boosted Trees") {
+    prm_ui_name <- params$tune_boost
+  } else if (method == "Cubist") {
+    prm_ui_name <- params$tune_cub
+  } else if (method == "Feed-Forward") {
+    prm_ui_name <- params$tune_ff
+  } else if (method == "Feed-Forward AR") {
+    prm_ui_name <- params$tune_ffar
+  } else if (method == "ARIMA-Boost") {
+    prm_ui_name <- params$tune_arima_boost
+  } else if (method == "Prophet-Boost") {
+    prm_ui_name <- params$tune_prophet_boost
   } else {
     stop(paste("Unknown method", method))
   }
   tune_params <- mtd_params[names(mtd_params) %in% prm_ui_name] # get the parameters to tune
   is_to_tune <- mtd_params %in% tune_params
-  new_params <- purrr::map2(mtd_params, is_to_tune, get_tune) |> purrr::set_names(mtd_params)
-  # update_params <- c(params, new_params)
+  new_params <- purrr::map2(mtd_params, is_to_tune, set_tune) |> purrr::set_names(mtd_params)
+
+  if (method == "SVM") {
+    new_params$boundary <- params$tune_boundary
+  }
+  if (method == "Boosted Trees") {
+    new_params$boost_method <- params$tune_boost_method
+  }
 
   return(new_params)
 
@@ -503,7 +529,7 @@ fit_model_tuning <- function(
     seed = 1992
 ) {
 
-  params_new <- update_tune_model_parameters(method, params)
+  params_new <- set_tune_parameters(method, params)
   check_parameters(method, params_new)
   set.seed(seed)
 
@@ -529,11 +555,8 @@ fit_model_tuning <- function(
   # grid_spec <- generate_grid_spec(method, model_spec, grid_size, seed)
 
   # tuning
-  if (n_folds > 10 | grid_size > 25) {
-    doFuture::registerDoFuture()
-    future::plan(strategy = "multisession", workers = parallelly::availableCores() - 1)
-    message("Number of parallel workers: ", future::nbrOfWorkers())
-  }
+  doFuture::registerDoFuture()
+  future::plan(strategy = "multisession", workers = parallelly::availableCores() - 1)
   tune_fit <- wkfl_spec |>
     tune::tune_grid(
       resamples = cv_splits,

dashboard/todo.txt

@@ -13,11 +13,7 @@ Next steps:
 - documentazione in alto a destra
 
 To Do:
-- fare check senza if else in update_tune_model_parameters
 - tune_bayes
-- finire update_tune_model_parameters con mapping parametri (occhio, non mtry ma rf_mtry)
-- testare flusso optimize fino al forecast
-- cambiare assegnazione nomi ai parametri in UI
 
 - aggiungere metodi di automl (h2o)
 - pensare e aggiungere la sezione di stacking (LM + Elastic Net)
@@ -26,3 +22,4 @@ To Do:
 - aggiungere metodi di dl (NeuralProphet + NBEATS + DeepAR)
 - pensare e aggiungere la sezione di feature engineering (con in mente il save)
 - modificare output modello con parsing
+- cambiare assegnazione nomi ai parametri in UI

dashboard/tsf_dashboard.Rmd

@@ -777,7 +777,7 @@ conditionalPanel(
 # Coming Soon!
 conditionalPanel(
   condition = "input.method == 'COMING SOON!'",
-  h5("New Deep Learning algorithms will be released soon!")
+  h5("New algorithms will be released soon!")
 )
 ```
 
@@ -1077,7 +1077,7 @@ dropdownButton(
 pickerInput(
   inputId = "tune_method", label = h3("Forecast Algorithm"), multiple = FALSE,
   choices = list(
-    `Machine Learning` = ml_methods,
+    `Machine Learning` = ml_methods[-1],
     `Deep Learning` = dl_methods,
     `Mixed Algorithms` = mix_methods
   ), 
@@ -1110,24 +1110,157 @@ conditionalPanel(
   condition = "input.tune_method == 'Elastic Net'",
   h5("Algorithm hyperparameters to optimize: "),
   pickerInput(
-    inputId = "tune_xx_elanet", label = NULL, # _xx_ needed to recognize it easily in update_tune_model_parameters()
+    inputId = "tune_elanet", label = NULL, 
     choices = mtd_prm_names[["Elastic Net"]], multiple = TRUE, 
     selected = mtd_prm_names[["Elastic Net"]],
     options = list("actions-box" = TRUE)
   )
 )
 
+# MARS
+conditionalPanel(
+  condition = "input.tune_method == 'MARS'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_mars", label = NULL, 
+    choices = mtd_prm_names[["MARS"]], multiple = TRUE, 
+    selected = mtd_prm_names[["MARS"]][1:2],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# KNN
+conditionalPanel(
+  condition = "input.tune_method == 'KNN'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_knn", label = NULL, 
+    choices = mtd_prm_names[["KNN"]], multiple = TRUE, 
+    selected = mtd_prm_names[["KNN"]],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# SVM
+conditionalPanel(
+  condition = "input.tune_method == 'SVM'",
+  h5("Algorithm hyperparameters to optimize: "),
+  prettyRadioButtons(
+    inputId = "tune_boundary", label = "Boundary Type", choices = c("Linear", "Radial"), 
+    inline = TRUE, selected = get_default("boundary")
+  ),
+  conditionalPanel(
+    condition = "input.tune_boundary == 'Linear'",
+    pickerInput(
+      inputId = "tune_svm_linear", label = NULL, multiple = TRUE,
+      choices = mtd_prm_names[["SVM"]][-c(1, 4)], selected = mtd_prm_names[["SVM"]][2:3],
+      options = list("actions-box" = TRUE)
+    )
+  ),
+  conditionalPanel(
+    condition = "input.tune_boundary == 'Radial'",
+    pickerInput(
+      inputId = "tune_svm_rbf", label = NULL, multiple = TRUE,
+      choices = mtd_prm_names[["SVM"]][-1], selected = mtd_prm_names[["SVM"]][2:3],
+      options = list("actions-box" = TRUE)
+    )
+  )
+)
+
 # Random Forest
 conditionalPanel(
   condition = "input.tune_method == 'Random Forest'",
   h5("Algorithm hyperparameters to optimize: "),
   pickerInput(
-    inputId = "tune_xx_rf", label = NULL, # _xx_ needed to recognize it easily in update_tune_model_parameters()
+    inputId = "tune_rf", label = NULL, 
     choices = mtd_prm_names[["Random Forest"]], multiple = TRUE, 
     selected = mtd_prm_names[["Random Forest"]],
     options = list("actions-box" = TRUE)
   )
 )
+
+# Boosted Trees
+conditionalPanel(
+  condition = "input.tune_method == 'Boosted Trees'",
+  h5("Algorithm hyperparameters to optimize: "),
+  prettyRadioButtons(
+    inputId = "tune_boost_method", label = "Boosting Method", 
+    choices = c("XGBoost", "LightGBM"), 
+    inline = TRUE, selected = get_default("boost_method")
+  ),
+  pickerInput(
+    inputId = "tune_boost", label = NULL, 
+    choices = mtd_prm_names[["Boosted Trees"]][-1], multiple = TRUE, 
+    selected = mtd_prm_names[["Boosted Trees"]][6],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# Cubist
+conditionalPanel(
+  condition = "input.tune_method == 'Cubist'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_cub", label = NULL, 
+    choices = mtd_prm_names[["Cubist"]], multiple = TRUE, 
+    selected = mtd_prm_names[["Cubist"]],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# Feed-Forward
+conditionalPanel(
+  condition = "input.tune_method == 'Feed-Forward'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_ff", label = NULL, 
+    choices = mtd_prm_names[["Feed-Forward"]], multiple = TRUE, 
+    selected = mtd_prm_names[["Feed-Forward"]][1:3],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# Feed-Forward AR
+conditionalPanel(
+  condition = "input.tune_method == 'Feed-Forward AR'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_ffar", label = NULL, 
+    choices = mtd_prm_names[["Feed-Forward AR"]], multiple = TRUE, 
+    selected = mtd_prm_names[["Feed-Forward AR"]][2:4],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# ARIMA-Boost
+conditionalPanel(
+  condition = "input.tune_method == 'ARIMA-Boost'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_arima_boost", label = NULL, 
+    choices = mtd_prm_names[["ARIMA-Boost"]], multiple = TRUE, 
+    selected = mtd_prm_names[["ARIMA-Boost"]][5],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# Prophet-Boost
+conditionalPanel(
+  condition = "input.tune_method == 'Prophet-Boost'",
+  h5("Algorithm hyperparameters to optimize: "),
+  pickerInput(
+    inputId = "tune_prophet_boost", label = NULL, 
+    choices = mtd_prm_names[["Prophet-Boost"]], multiple = TRUE, 
+    selected = mtd_prm_names[["Prophet-Boost"]][5],
+    options = list("actions-box" = TRUE)
+  )
+)
+
+# Coming Soon!
+conditionalPanel(
+  condition = "input.tune_method == 'COMING SOON!'",
+  h5("New algorithms will be released soon!")
+)
 ```
 
 ```{r}