Merge pull request #164 from microsoft/mitokic/07292024/multistep-bug

Mitokic/07292024/multistep bug

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: finnts
 Title: Microsoft Finance Time Series Forecasting Framework
-Version: 0.4.0.9005
+Version: 0.4.0.9006
 Authors@R: 
     c(person(given = "Mike",
            family = "Tokic",

NEWS.md

@@ -1,4 +1,4 @@
-# finnts 0.4.0.9005 (DEVELOPMENT VERSION)
+# finnts 0.4.0.9006 (DEVELOPMENT VERSION)
 
 ## Improvements
 
@@ -8,6 +8,7 @@
 -   Always save the most accurate model average, regardless if selected as best model. This allows for improved scaling with large data sets.
 -   Automatically condense large forecasts (+10k time series) into smaller amount of files to make it easier to read forecast outputs
 -   Improved weighted MAPE calculation across all time series
+-   Changed default for box_cox argument in `prep_data()` to FALSE
 
 ## Bug Fixes
 

R/ensemble_models.R

@@ -366,6 +366,7 @@ ensemble_models <- function(run_info,
             parallel_over = "everything"
           )
         ) %>%
+          base::suppressMessages() %>%
           base::suppressWarnings()
 
         best_param <- tune::select_best(tune_results, metric = "rmse")
@@ -397,7 +398,9 @@ ensemble_models <- function(run_info,
             pkgs = inner_packages,
             parallel_over = "everything"
           )
-        )
+        ) %>%
+          base::suppressMessages() %>%
+          base::suppressWarnings()
 
         final_fcst <- tune::collect_predictions(refit_tbl) %>%
           dplyr::rename(

R/multistep_cubist.R

@@ -469,8 +469,8 @@ cubist_multistep_predict_impl <- function(object, new_data, ...) {
 
     xreg_tbl_final <- xreg_tbl %>%
       dplyr::filter(
-        Run_Number >= start_val,
-        Run_Number <= lag_number
+        Run_Number >= as.numeric(start_val),
+        Run_Number <= as.numeric(lag_number)
       )
 
     if (!is.null(xreg_tbl)) {

R/multistep_glmnet.R

@@ -457,8 +457,8 @@ glmnet_multistep_predict_impl <- function(object, new_data, ...) {
 
     xreg_tbl_final <- xreg_tbl %>%
       dplyr::filter(
-        Run_Number >= start_val,
-        Run_Number <= lag_number
+        Run_Number >= as.numeric(start_val),
+        Run_Number <= as.numeric(lag_number)
       )
 
     if (!is.null(xreg_tbl)) {

R/multistep_mars.R

@@ -480,8 +480,8 @@ mars_multistep_predict_impl <- function(object, new_data, ...) {
 
     xreg_tbl_final <- xreg_tbl %>%
       dplyr::filter(
-        Run_Number >= start_val,
-        Run_Number <= lag_number
+        Run_Number >= as.numeric(start_val),
+        Run_Number <= as.numeric(lag_number)
       )
 
     if (!is.null(xreg_tbl)) {

R/multistep_svm_poly.R

@@ -506,8 +506,8 @@ svm_poly_multistep_predict_impl <- function(object, new_data, ...) {
 
     xreg_tbl_final <- xreg_tbl %>%
       dplyr::filter(
-        Run_Number >= start_val,
-        Run_Number <= lag_number
+        Run_Number >= as.numeric(start_val),
+        Run_Number <= as.numeric(lag_number)
       )
 
     if (!is.null(xreg_tbl)) {

R/multistep_svm_rbf.R

@@ -486,8 +486,8 @@ svm_rbf_multistep_predict_impl <- function(object, new_data, ...) {
 
     xreg_tbl_final <- xreg_tbl %>%
       dplyr::filter(
-        Run_Number >= start_val,
-        Run_Number <= lag_number
+        Run_Number >= as.numeric(start_val),
+        Run_Number <= as.numeric(lag_number)
       )
 
     if (!is.null(xreg_tbl)) {

R/multistep_xgboost.R

@@ -568,8 +568,8 @@ xgboost_multistep_predict_impl <- function(object, new_data, ...) {
 
     xreg_tbl_temp <- xreg_tbl %>%
       dplyr::filter(
-        Run_Number >= start_val,
-        Run_Number <= lag_number
+        Run_Number >= as.numeric(start_val),
+        Run_Number <= as.numeric(lag_number)
       )
 
     xreg_tbl_final <- xreg_tbl_temp %>%

R/prep_data.R

@@ -83,7 +83,7 @@ prep_data <- function(run_info,
                       fiscal_year_start = 1,
                       clean_missing_values = TRUE,
                       clean_outliers = FALSE,
-                      box_cox = TRUE,
+                      box_cox = FALSE,
                       stationary = TRUE,
                       forecast_approach = "bottoms_up",
                       parallel_processing = NULL,

R/prep_models.R

@@ -804,7 +804,9 @@ model_hyperparameters <- function(run_info,
 
   # update logging file
   log_df <- log_df %>%
-    dplyr::mutate(num_hyperparameters = num_hyperparameters)
+    dplyr::mutate(
+      num_hyperparameters = num_hyperparameters
+    )
 
   write_data(
     x = log_df,

R/read_write_data.R

@@ -407,10 +407,17 @@ write_data <- function(x,
 write_data_type <- function(x,
                             path,
                             type) {
+  if (type == "csv") {
+    if (nrow(x) == 1) {
+      type <- "log"
+    }
+  }
+
   switch(type,
     rds = saveRDS(x, path),
     parquet = arrow::write_parquet(x, path),
     csv = vroom::vroom_write(x, path, delim = ",", progress = FALSE),
+    log = utils::write.csv(x, path, row.names = FALSE),
     qs = qs::qsave(x, path)
   )
 }

R/train_models.R

@@ -516,7 +516,9 @@ train_models <- function(run_info,
             parallel_over = "everything"
           )
         ) %>%
-          tune::collect_predictions()
+          tune::collect_predictions() %>%
+          base::suppressMessages() %>%
+          base::suppressWarnings()
 
         # finalize forecast
         final_fcst <- refit_tbl %>%
@@ -534,6 +536,11 @@ train_models <- function(run_info,
           dplyr::mutate(Hyperparameter_ID = hyperparameter_id) %>%
           dplyr::select(-.row, -.config)
 
+        # check for future forecast
+        if (as.numeric(min(unique(final_fcst$Train_Test_ID))) != 1) {
+          stop("model is missing future forecast")
+        }
+
         # undo differencing transformation
         if (stationary & model %in% list_multivariate_models()) {
           if (combo_hash == "All-Data") {