nk773

generate <- function(mu=0, sigma=1, dist="normal", nocols=5, norows=100, min=0, max=1, sd=3000) {
  set.seed(sd)
  if (dist == "uniform") {
    df <- replicate(nocols,runif(norows)*(min+(max-min)))
    return(df)
  }
  if (dist == "normal") {
    df <- replicate(nocols,rnorm(norows, mu, sigma)*(min+(max-min)))
    return (df)
  }

nk773

concatenate<-function(dataset1, colrng, sep=" ") {
  
  rng <- which(names(dataset1)%in%colrng)
  if (length(rng)<2) {
    print(paste("Insufficient columns for concatenation", length(rng), rng))
    return(dataset1)
  }
  df<-dataset1[,-rng]
  
  df3<-apply(dataset1[,rng],1, paste, collapse=sep)

nk773

addMeanFeature<-function(dataset1, colrng, bycol=NA, windowsize){
  library(zoo)
  id<-bycol
  #print(paste("id =", id))
  rollingmean = c()
  rollingsd = c()
  rng <- which(names(dataset1)%in%colrng)
  a = paste("a",(1:length(rng)),sep="") # average
  sd =paste("sd",(1:length(rng)),sep="") # standard deviation
  

nk773

TextToCols<-function(dataset1, dataset2 = NA, selcol, splch, incorig) {
  if (length(selcol)>1)
  {
    
    scol<-selcol[1]
  }
  else
  {
    scol<-selcol
  }

nk773

# Map 1-based optional input ports to variables
library(forecast)
masefun <- function(observed, predicted){
  error = 0;
  if (length(observed) != length(predicted)) {
    return (NA);
  } else if (length(observed) == 0 || length(predicted) == 0) {
    return (NA);
  }
  else {

nk773

fsscore<-function(model, numPeriodsToForecast) {
  if (numPeriodsToForecast<=0)
  {
    print("ERROR: forecast doesn't have any unknow value as time attribute present in training data")
    return(data.frame(NA))
  }
  else
  {
   
    forecastedData <- forecast(model, h=numPeriodsToForecast)

nk773

compareData<-function(dataset1, dataset2)
{
  # compares two datasets for each cell value. If dimensions of dataset differ, it uses minimum dimension
  # Args:
  #   dataset1: first dataset to be compared.
  #   dataset2: second dataset to be compared.
  #
  # Returns:
  #   Returns list of the following
  #       dataframe containing boolean matrix for each cell indicating match or no match

nk773

lf2 <- function (dataset2, valcol2, numfeats, skps){
  # Computes the lag features for specific column specified by valcol2 within dataset2
  # This creates total numfeats new features and it assumes dataset is already sorted by time
  # Args:
  #   dataset2: dataset to be used for adding new features.
  #   valcol2:  index of the column for which we need to create new features.
  #   numfeats: number of features to create. For example, if this is 3, then it will create 3 new lag features
  #   skps:     created nested lag features by using this array as skipping interval for each level
  # Returns:
  #   New dataset with additional features

nk773

lf1 <- function (dataset2, valcol2, numfeats){
  # Computes the lag features for specific column specified by valcol2 within dataset2
  # This creates total numfeats new features and it assumes dataset is already sorted by time
  # Args:
  #   dataset2: dataset to be used for adding new features.
  #   valcol2:  index of the column for which we need to create new features.
  #   numfeats: number of features to create. For example, if this is 3, then it will create 3 new lag features
  #
  # Returns:
  #   New dataset with additional features

nk773

#!/bin/bash

if [ $# -lt 3 ]
then
   echo "Number of arguments doesn't have required mandatory parameters"
   echo "azureml_rrs <PathToRRSjsonFile> <API key> <RRS URL>"
   echo "Path to RRS json file    $1"
   echo "API Key                  $2"
   echo "RRS URL                  $3"
   echo "Total Arguments          $#"