fix clean calculation

business-science · Oct 29, 2023 · d9a251c · d9a251c
1 parent 1f04fab
commit d9a251c
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 7 deletions.
diff --git a/R/00_global_vars.R b/R/00_global_vars.R
@@ -10,5 +10,5 @@ globalVariables(
       ".metric", ".metrics", ".rank_metric", ".rank_std_err", "failure_rate", "std_err",
       ".rank_failure_rate", ".txt", "training", "testing", "fitted", "model", "..val", "..nm",
       ".white_noise_lower", ".white_noise_upper", ".date_var_collapsed", ".rowid", "..date_agg", "grp_names",
-      "check_row_exists", ".box_group", ".observed_cleaned")
+      "check_row_exists", ".box_group", ".observed_clean")
 )
diff --git a/R/anomalize-plotting.R b/R/anomalize-plotting.R
@@ -713,7 +713,7 @@ plot_anomalies_cleaned <- function(
     }
 
     column_names <- names(.data)
-    check_names  <- c("observed", "observed_cleaned") %in% column_names
+    check_names  <- c("observed", "observed_clean") %in% column_names
     if (!all(check_names)) stop('Error in plot_anomalies_decomp(): column names are missing. Run `anomalize()` and make sure: observed, remainder, anomaly, recomposed_l1, and recomposed_l2 are present', call. = FALSE)
 
     UseMethod("plot_anomalies_cleaned", .data)
@@ -821,10 +821,10 @@ plot_anomalies_cleaned.data.frame <- function(
     g <- g +
         ggplot2::scale_color_manual(values = c(.line_color, .cleaned_line_color))
 
-    # Add line - observed_cleaned
+    # Add line - observed_clean
     g <- g +
         ggplot2::geom_line(
-            ggplot2::aes(y = observed_cleaned, color = "Observed Cleaned"),
+            ggplot2::aes(y = observed_clean, color = "Observed Cleaned"),
             # color     = .cleaned_line_color,
             linewidth = .cleaned_line_size,
             linetype  = .cleaned_line_type,

diff --git a/R/anomalize.R b/R/anomalize.R
@@ -190,9 +190,9 @@ anomalize.data.frame <- function(
     # Clean
     ret <- ret %>%
         dplyr::mutate(
-            observed_cleaned = dplyr::case_when(
-                anomaly_direction == -1 ~ .clean_alpha * recomposed_l1,
-                anomaly_direction == 1 ~ .clean_alpha * recomposed_l2,
+            observed_clean = dplyr::case_when(
+                anomaly_direction == -1 ~ recomposed_l1 + (1-.clean_alpha)*anomaly_score,
+                anomaly_direction == 1 ~ recomposed_l2 - (1-.clean_alpha)*anomaly_score,
                 TRUE ~ observed
             )
         )