FakeDataGenerator: FakeDataGenerator

In AdrianAntico/RemixAutoML: AutoQuant

FakeDataGenerator

Description

Create fake data for examples

Usage

FakeDataGenerator(
  Correlation = 0.7,
  N = 1000L,
  ID = 5L,
  FactorCount = 2L,
  AddDate = TRUE,
  AddComment = FALSE,
  AddWeightsColumn = FALSE,
  ZIP = 5L,
  TimeSeries = FALSE,
  TimeSeriesTimeAgg = "day",
  ChainLadderData = FALSE,
  Classification = FALSE,
  MultiClass = FALSE
)

Arguments

Correlation	Set the correlation value for simulated data
N	Number of records
ID	Number of IDcols to include
FactorCount	Number of factor type columns to create
AddDate	Set to TRUE to include a date column
AddComment	Set to TRUE to add a comment column
ZIP	Zero Inflation Model target variable creation. Select from 0 to 5 to create that number of distinctly distributed data, stratifed from small to large
TimeSeries	For testing AutoBanditSarima
TimeSeriesTimeAgg	Choose from "1min", "5min", "10min", "15min", "30min", "hour", "day", "week", "month", "quarter", "year",
ChainLadderData	Set to TRUE to return Chain Ladder Data for using AutoMLChainLadderTrainer
Classification	Set to TRUE to build classification data
MultiClass	Set to TRUE to build MultiClass data

Author(s)

Adrian Antico

Examples

## Not run: 
# Create dummy data to test regression, classification, and multiclass models.
#   I don't care too much about actual relationships but I can test out on the
#   regression problem since those variables will be correlated. The binary and
#   multiclass won't however since they were created separately.

# Regression
data <- AutoQuant::FakeDataGenerator(
  Correlation = 0.77,
  N = 1000000L,
  ID = 4L,
  FactorCount = 5L,
  AddDate = TRUE,
  AddComment = TRUE,
  AddWeightsColumn = TRUE,
  ZIP = 0L,
  TimeSeries = FALSE,
  TimeSeriesTimeAgg = "day",
  ChainLadderData = FALSE,
  Classification = FALSE,
  MultiClass = FALSE)

# Classification
data2 <- AutoQuant::FakeDataGenerator(
  Correlation = 0.77,
  N = 1000000L,
  ID = 4L,
  FactorCount = 5L,
  AddDate = TRUE,
  AddComment = TRUE,
  AddWeightsColumn = TRUE,
  ZIP = 0L,
  TimeSeries = FALSE,
  TimeSeriesTimeAgg = "day",
  ChainLadderData = FALSE,
  Classification = TRUE,
  MultiClass = FALSE)

# MultiClass
data3 <- AutoQuant::FakeDataGenerator(
  Correlation = 0.77,
  N = 1000000L,
  ID = 4L,
  FactorCount = 5L,
  AddDate = TRUE,
  AddComment = TRUE,
  AddWeightsColumn = TRUE,
  ZIP = 0L,
  TimeSeries = FALSE,
  TimeSeriesTimeAgg = "day",
  ChainLadderData = FALSE,
  Classification = FALSE,
  MultiClass = TRUE)

data.table::setnames(data, 'Adrian', 'RegressionTarget')
data.table::setnames(data2, 'Adrian', 'BinaryTarget')
data.table::setnames(data3, 'Adrian', 'MultiClassTarget')

data <- cbind(data, data2$BinaryTarget, data3$MultiClassTarget)
data.table::setnames(data, c('V2','V3'), c('BinaryTarget','MultiClassTarget'))
data.table::setcolorder(data, c(1, c(ncol(data)-1,ncol(data),2:(ncol(data)-2))))

# Load to warehouse
AutoQuant::PostGRE_RemoveCreateAppend(
  data = data,
  Append = TRUE,
  TableName = "App_QA_BigData",
  CloseConnection = TRUE,
  CreateSchema = NULL,
  Host = "localhost",
  DBName = "AutoQuant",
  User = "postgres",
  Port = 5432,
  Password = "",
  Temporary = FALSE,
  Connection = NULL)

## End(Not run)

AdrianAntico/RemixAutoML documentation built on Feb. 3, 2024, 3:32 a.m.