aggregate

R/aggregate.R

@@ -0,0 +1,43 @@
+#' Aggregate predictions
+#' 
+#' This function aggregates predictions from multiple
+#' calls of `prediction()`.
+#'
+#' @param predictions (vector of predict.rfsrc() objects) The prediction objects from `prediction()`.
+#' 
+#' @return something 
+#' 
+#' @export
+#'
+#' @examples 
+#' \dontrun{
+#' # Example usage:
+#' prediction <- list()
+#' predictions[[1]] <- prediction(model = "bs23.rds", ...)
+#' predictions[[2]] <- prediction(model = "bs24.rds", ...)
+#' result <- aggregate(predictions)
+#' }
+#' 
+#' @importFrom logger log_info
+#' @importFrom utils read.csv write.table
+#' @importFrom magrittr `%>%`
+#' @importFrom randomForestSRC predict.rfsrc
+
+aggregate <- function(predictions) {
+
+  # Get all the confidence scores
+  scores <- lapply(predictions, function(pred) pred$predicted)
+
+  # Get averages of different sources
+  all_scores <- do.call(rbind, scores)
+
+  # Calculate the averages and standard deviations for each category
+  category_summary <- apply(all_scores, 2, function(x) c(mean = mean(x), sd = sd(x)))
+
+# TODO return averaged VIMP list
+# TODO make the return variable an object holding all this stuff
+# TODO document this better
+
+  return(category_summary)
+}
+

tests/testthat/test-20_aggregation.R

@@ -0,0 +1,38 @@
+suppressPackageStartupMessages(library("data.table"))
+suppressPackageStartupMessages(library("dplyr"))
+suppressPackageStartupMessages(library("logger"))
+suppressPackageStartupMessages(library("gt"))
+suppressPackageStartupMessages(library("randomForestSRC"))
+suppressPackageStartupMessages(library("tidyverse"))
+
+
+test_that("Aggregating LMO0003 with example_query", {
+
+  ncores <- 1
+
+  # rfsrc prediction objects
+  predictions <- list()
+  for(i in seq(1,3)){
+    model <- paste0("test-results/bs",(i+22),".rds")
+    predictions[[i]] <- prediction(model_filename = model, 
+                                   query = "example_query.csv", ncores = ncores)
+
+  }
+
+  bootstrapped_prediction <- aggregate(predictions = predictions)
+
+  expected <- matrix(
+    c(
+      0.1501267932, 0.1578980673, 0.3111404334, 0.1697799469, 0.211054759,
+      0.0006860643, 0.0008925544, 0.0002023788, 0.0009216065, 0.000183757
+    ),
+    nrow = 2, byrow = TRUE,
+    dimnames = list(c("mean", "sd"), c("dairy", "fruit", "meat", "seafood", "vegetable"))
+  )
+
+  expect_equal(bootstrapped_prediction["mean","dairy"], 
+               expected["mean","dairy"], 
+               tolerance = 0.1)
+  expect_equal(bootstrapped_prediction, expected, tolerance = 0.1)
+
+})