Merge pull request #3 from hendersontrent/trent-dev

maxabs scaling

NAMESPACE

@@ -1,5 +1,6 @@
 # Generated by roxygen2: do not edit by hand
 
+export(maxabs_scaler)
 export(minmax_scaler)
 export(normalise)
 export(normalize)

R/normalise.R

@@ -6,14 +6,14 @@
 #' @importFrom scales rescale
 #' @importFrom dplyr %>% group_by mutate ungroup
 #' @param data either a \code{feature_calculations} object containing the raw feature matrix produced by \code{calculate_features} or a \code{vector} of class \code{numeric} containing values to be rescaled
-#' @param norm_method \code{character} denoting the rescaling/normalising method to apply. Can be one of \code{"zScore"}, \code{"Sigmoid"}, \code{"RobustSigmoid"}, or \code{"MinMax"}. Defaults to \code{"zScore"}
+#' @param norm_method \code{character} denoting the rescaling/normalising method to apply. Can be one of \code{"zScore"}, \code{"Sigmoid"}, \code{"RobustSigmoid"}, \code{"MinMax"}, or \code{"MaxAbs"}. Defaults to \code{"zScore"}
 #' @param unit_int \code{Boolean} whether to rescale into unit interval \code{[0,1]} after applying normalisation method. Defaults to \code{FALSE}
 #' @return either an object of class \code{data.frame} or a \code{numeric} vector
 #' @author Trent Henderson
 #' @export
 #'
 
-normalise <- function(data, norm_method = c("zScore", "Sigmoid", "RobustSigmoid", "MinMax"), unit_int = FALSE){
+normalise <- function(data, norm_method = c("zScore", "Sigmoid", "RobustSigmoid", "MinMax", "MaxAbs"), unit_int = FALSE){
 
   norm_method <- match.arg(norm_method)
 
@@ -46,6 +46,11 @@ normalise <- function(data, norm_method = c("zScore", "Sigmoid", "RobustSigmoid"
         dplyr::mutate(values = minmax_scaler(.data$values))
     }
 
+    if(norm_method == "MaxAbs"){
+      normed <- normed %>%
+        dplyr::mutate(values = maxabs_scaler(.data$values))
+    }
+
     if(unit_int){
       normed <- normed %>%
         dplyr::mutate(values = scales::rescale(.data$values, to = c(0, 1)))
@@ -74,6 +79,10 @@ normalise <- function(data, norm_method = c("zScore", "Sigmoid", "RobustSigmoid"
       normed <- minmax_scaler(data)
     }
 
+    if(norm_method == "MaxAbs"){
+      normed <- maxabs_scaler(data)
+    }
+
     if(unit_int){
       normed <- scales::rescale(normed, to = c(0, 1))
     }

R/rescalers.R

@@ -66,3 +66,19 @@ robustsigmoid_scaler <- function(x){
   x_new <- 1 / (1 + exp(-((x1 - stats::median(x1, na.rm = TRUE)) / (stats::IQR(x1, na.rm = TRUE) / 1.35))))
   return(x_new)
 }
+
+#' Rescales a numeric vector using maximum absolute scaling
+#'
+#' \eqn{z_{i} = \frac{x_{i}}{\text{max}(\mathbf{x})}}
+#'
+#' @param x \code{numeric} vector
+#' @return \code{numeric} vector
+#' @author Trent Henderson
+#' @export
+#'
+
+maxabs_scaler <- function(x){
+  x1 <- as.vector(x) # Catches class "ts" cases
+  x_new <- x1 / max(x1, na.rm = TRUE)
+  return(x_new)
+}

README.Rmd

@@ -23,9 +23,10 @@ devtools::install_github("hendersontrent/normaliseR")
 
 `normaliseR` is a software package for R for rescaling numerical vectors or `feature_calculations` objects produced by the [`purloiner`](https://github.com/hendersontrent/purloiner) package. It was specifically designed to be the rescaling module of the broader feature-based time-series 'mothership' package [`theft`](https://github.com/hendersontrent/theft).
 
-Putting calculated feature vectors on an equal scale is crucial for any statistical or machine learning model as variables with high variance can adversely impact the model's capacity to fit the data appropriately, learn appropriate weight values, or minimise a loss function. `normaliseR` includes function `normalise` (or `normalize`) to rescale either a whole `feature_calculations` object, or a single vector of values. Four normalisation methods are currently offered:
+Putting calculated feature vectors on an equal scale is crucial for any statistical or machine learning model as variables with high variance can adversely impact the model's capacity to fit the data appropriately, learn appropriate weight values, or minimise a loss function. `normaliseR` includes function `normalise` (or `normalize`) to rescale either a whole `feature_calculations` object, or a single vector of values. The following normalisation methods are currently offered:
 
 * z-score---`"zScore"`
 * Sigmoid---`"Sigmoid"`
 * Outlier-robust Sigmoid (credit to Ben Fulcher for creating the original [MATLAB version](https://github.com/benfulcher/hctsa)) -- `"RobustSigmoid"`
 * Min-max---`"MinMax"`
+* Maximum absolute---`"MaxAbs"`

README.md

@@ -27,11 +27,12 @@ can adversely impact the model’s capacity to fit the data appropriately,
 learn appropriate weight values, or minimise a loss function.
 `normaliseR` includes function `normalise` (or `normalize`) to rescale
 either a whole `feature_calculations` object, or a single vector of
-values. Four normalisation methods are currently offered:
+values. The following normalisation methods are currently offered:
 
 - z-score—`"zScore"`
 - Sigmoid—`"Sigmoid"`
 - Outlier-robust Sigmoid (credit to Ben Fulcher for creating the
   original [MATLAB version](https://github.com/benfulcher/hctsa)) –
   `"RobustSigmoid"`
 - Min-max—`"MinMax"`
+- Maximum absolute—`"MaxAbs"`

docs/pkgdown.yml

@@ -3,7 +3,7 @@ pkgdown: 2.0.7
 pkgdown_sha: ~
 articles:
   normaliseR: normaliseR.html
-last_built: 2023-08-20T10:34Z
+last_built: 2023-08-20T11:40Z
 urls:
   reference: https://hendersontrent.github.io/normaliseR/reference
   article: https://hendersontrent.github.io/normaliseR/articles