feat: add properties to forecast learners

m-muecke · m-muecke · commit cc9cdfd4589c · 2025-01-08T21:05:44.000+01:00
diff --git a/R/LearnerRegrArfima.R b/R/LearnerRegrArfima.R
@@ -27,6 +27,7 @@ LearnerFcstArfima = R6Class("LearnerFcstArfima",
         param_set = param_set,
         predict_types = c("response", "quantiles"),
         feature_types = c("Date", "logical", "integer", "numeric"),
+        properties = c("univariate", "exogenous", "missings"),
         packages = c("mlr3forecast", "forecast"),
         label = "ARFIMA",
         man = "mlr3forecast::mlr_learners_fcst.arfima"
diff --git a/R/LearnerRegrArima.R b/R/LearnerRegrArima.R
@@ -39,6 +39,7 @@ LearnerFcstArima = R6Class("LearnerFcstArima",
         param_set = param_set,
         predict_types = c("response", "quantiles"),
         feature_types = c("Date", "logical", "integer", "numeric"),
+        properties = c("univariate", "exogenous", "missings"),
         packages = c("mlr3forecast", "forecast"),
         label = "ARIMA",
         man = "mlr3forecast::mlr_learners_fcst.arima"
diff --git a/R/LearnerRegrAutoArima.R b/R/LearnerRegrAutoArima.R
@@ -1,4 +1,4 @@
-#' @title ARIMA
+#' @title Auto ARIMA
 #'
 #' @name mlr_learners_fcst.auto_arima
 #'
@@ -44,6 +44,7 @@ LearnerFcstAutoArima = R6Class("LearnerFcstAutoArima",
         param_set = param_set,
         predict_types = c("response", "quantiles"),
         feature_types = c("Date", "logical", "integer", "numeric"),
+        properties = c("univariate", "exogenous", "missings"),
         packages = c("mlr3forecast", "forecast"),
         label = "Auto ARIMA",
         man = "mlr3forecast::mlr_learners_fcst.arima"
diff --git a/R/LearnerRegrEts.R b/R/LearnerRegrEts.R
@@ -46,6 +46,7 @@ LearnerFcstEts = R6Class("LearnerFcstEts",
         param_set = param_set,
         predict_types = c("response", "quantiles"),
         feature_types = c("Date", "logical", "integer", "numeric"),
+        properties = c("univariate", "missings"),
         packages = c("mlr3forecast", "forecast"),
         label = "ETS",
         man = "mlr3forecast::mlr_learners_fcst.ets"
diff --git a/R/zzz.R b/R/zzz.R
@@ -12,6 +12,7 @@ mlr3forecast_learners = new.env()
 mlr3forecast_measures = new.env()
 mlr3forecast_feature_types = c(dte = "Date")
 mlr3forecast_col_roles = "key"
+mlr3forecast_learner_properties = c("univariate", "multivariate", "exogenous", "missings")
 
 named_union = function(x, y) set_names(union(x, y), union(names(x), names(y)))
 
@@ -36,7 +37,12 @@ register_mlr3 = function() {
     "fcst", "mlr3forecast", "TaskRegr", "LearnerRegr", "PredictionFcst", "PredictionDataFcst", "MeasureFcst" # nolint
   ), fill = TRUE), "type")
   mlr_reflections$learner_predict_types$fcst = mlr_reflections$learner_predict_types$regr
-  mlr_reflections$learner_properties$fcst = mlr_reflections$learner_properties$regr
+  mlr_reflections$learner_properties$fcst = union(
+    mlr_reflections$learner_properties$regr, mlr3forecast_learner_properties
+  )
+  mlr_reflections$learner_properties$regr = union(
+    mlr_reflections$learner_properties$regr, mlr3forecast_learner_properties
+  )
   mlr_reflections$task_col_roles$fcst = mlr_reflections$task_col_roles$regr
   mlr_reflections$task_col_roles$regr = union(
     mlr_reflections$task_col_roles$regr, mlr3forecast_col_roles
diff --git a/README.md b/README.md
@@ -44,32 +44,32 @@ prediction = flrn$predict_newdata(newdata, task)
 prediction
 #> <PredictionRegr> for 3 observations:
 #>  row_ids truth response
-#>        1    NA 434.9580
-#>        2    NA 438.3391
-#>        3    NA 456.1386
+#>        1    NA 438.0155
+#>        2    NA 439.6429
+#>        3    NA 457.8119
 prediction = flrn$predict(task, 142:144)
 prediction
 #> <PredictionRegr> for 3 observations:
 #>  row_ids truth response
-#>        1   461 456.7837
-#>        2   390 412.4510
-#>        3   432 434.0057
+#>        1   461 459.7825
+#>        2   390 417.8945
+#>        3   432 435.8002
 prediction$score(msr("regr.rmse"))
 #> regr.rmse 
-#>  13.23942
+#>  16.26886
 
 flrn = ForecastLearner$new(lrn("regr.ranger"), 1:12)
 resampling = rsmp("forecast_holdout", ratio = 0.9)
 rr = resample(task, flrn, resampling)
 rr$aggregate(msr("regr.rmse"))
 #> regr.rmse 
-#>  47.49655
+#>  49.22924
 
 resampling = rsmp("forecast_cv")
 rr = resample(task, flrn, resampling)
 rr$aggregate(msr("regr.rmse"))
 #> regr.rmse 
-#>  25.05562
+#>  25.57887
 ```
 
 ### Multivariate
@@ -89,34 +89,34 @@ flrn = ForecastLearner$new(lrn("regr.ranger"), 1:12)$train(new_task)
 prediction = flrn$predict(new_task, 142:144)
 prediction$score(msr("regr.rmse"))
 #> regr.rmse 
-#>  13.13675
+#>  12.35229
 
 row_ids = new_task$nrow - 0:2
 flrn$predict_newdata(new_task$data(rows = row_ids), new_task)
 #> <PredictionRegr> for 3 observations:
 #>  row_ids truth response
-#>        1   432 435.4188
-#>        2   390 434.2164
-#>        3   461 457.9859
+#>        1   432 431.1496
+#>        2   390 429.3616
+#>        3   461 455.1301
 newdata = new_task$data(rows = row_ids, cols = new_task$feature_names)
 flrn$predict_newdata(newdata, new_task)
 #> <PredictionRegr> for 3 observations:
 #>  row_ids truth response
-#>        1    NA 435.4188
-#>        2    NA 434.2164
-#>        3    NA 457.9859
+#>        1    NA 431.1496
+#>        2    NA 429.3616
+#>        3    NA 455.1301
 
 resampling = rsmp("forecast_holdout", ratio = 0.9)
 rr = resample(new_task, flrn, resampling)
 rr$aggregate(msr("regr.rmse"))
 #> regr.rmse 
-#>  47.83001
+#>   48.2451
 
 resampling = rsmp("forecast_cv")
 rr = resample(new_task, flrn, resampling)
 rr$aggregate(msr("regr.rmse"))
 #> regr.rmse 
-#>  27.82117
+#>  26.80115
 ```
 
 ### mlr3pipelines integration
@@ -131,7 +131,7 @@ glrn = as_learner(graph %>>% flrn)$train(task)
 prediction = glrn$predict(task, 142:144)
 prediction$score(msr("regr.rmse"))
 #> regr.rmse 
-#>   11.0778
+#>  12.39349
 ```
 
 ### Example: Forecasting electricity demand
@@ -205,14 +205,14 @@ flrn = ForecastLearner$new(lrn("regr.ranger"), 1:3)$train(task)
 prediction = flrn$predict(task, 4460:4464)
 prediction$score(msr("regr.rmse"))
 #> regr.rmse 
-#>  25052.84
+#>  23543.57
 
 flrn = ForecastLearner$new(lrn("regr.ranger"), 1:3)
 resampling = rsmp("forecast_holdout", ratio = 0.9)
 rr = resample(task, flrn, resampling)
 rr$aggregate(msr("regr.rmse"))
 #> regr.rmse 
-#>  93423.87
+#>     92241
 ```
 
 ### Example: Global vs Local Forecasting
@@ -247,7 +247,7 @@ row_ids = tab[year >= 2015, row_id]
 prediction = flrn$predict(task, row_ids)
 prediction$score(msr("regr.rmse"))
 #> regr.rmse 
-#>  29931.59
+#>  32967.19
 
 # global forecasting
 task = tsibbledata::aus_livestock |>
@@ -268,7 +268,7 @@ row_ids = tab[year >= 2015 & state == "Western Australia", row_id]
 prediction = flrn$predict(task, row_ids)
 prediction$score(msr("regr.rmse"))
 #> regr.rmse 
-#>  31607.32
+#>  31955.28
 ```
 
 ### Example: generate new data
