AutoH2oGLMMultiClass: AutoH2oGLMMultiClass

Description

AutoH2oGLMMultiClass is an automated H2O modeling framework with grid-tuning and model evaluation that runs a variety of steps. First, a stratified sampling (by the target variable) is done to create train and validation sets. Then, the function will run a random grid tune over N number of models and find which model is the best (a default model is always included in that set). Once the model is identified and built, several other outputs are generated: validation data with predictions, evaluation metrics, confusion matrix, and variable importance.

Usage

AutoH2oGLMMultiClass(
  data,
  TrainOnFull = FALSE,
  ValidationData = NULL,
  TestData = NULL,
  TargetColumnName = NULL,
  FeatureColNames = NULL,
  RandomColNumbers = NULL,
  InteractionColNumbers = NULL,
  WeightsColumn = NULL,
  MaxMem = {     gc()    
    paste0(as.character(floor(as.numeric(system("awk '/MemFree/ {print $2}' /proc/meminfo",
    intern = TRUE))/1e+06)), "G") },
  NThreads = max(1, parallel::detectCores() - 2),
  ModelID = "FirstModel",
  ReturnModelObjects = TRUE,
  model_path = NULL,
  metadata_path = NULL,
  SaveModelObjects = FALSE,
  SaveInfoToPDF = FALSE,
  IfSaveModel = "mojo",
  H2OShutdown = TRUE,
  H2OStartUp = TRUE,
  TransformNumericColumns = NULL,
  Methods = c("YeoJohnson", "BoxCox", "Asinh", "Log", "LogPlus1", "Sqrt", "Asin",
    "Logit"),
  MaxModelsInGrid = 2,
  NumOfParDepPlots = 3,
  GridTune = FALSE,
  GridStrategy = "Cartesian",
  StoppingRounds = 10,
  MaxRunTimeSecs = 3600 * 24 * 7,
  Distribution = "multinomial",
  Link = "family_default",
  eval_metric = "logloss",
  RandomDistribution = NULL,
  RandomLink = NULL,
  Solver = "AUTO",
  Alpha = 0.5,
  Lambda = NULL,
  LambdaSearch = FALSE,
  NLambdas = -1,
  Standardize = TRUE,
  RemoveCollinearColumns = FALSE,
  InterceptInclude = TRUE,
  NonNegativeCoefficients = FALSE
)

Arguments

data

This is your data set for training and testing your model

TrainOnFull

Set to TRUE to train on full data

ValidationData

This is your holdout data set used in modeling either refine your hyperparameters.

TestData

This is your holdout data set. Catboost using both training and validation data in the training process so you should evaluate out of sample performance with this data set.

TargetColumnName

Either supply the target column name OR the column number where the target is located (but not mixed types).

FeatureColNames

Either supply the feature column names OR the column number where the target is located (but not mixed types)

RandomColNumbers

Random effects column number indicies

InteractionColNumbers

Column numbers of the features you want to be pairwise interacted

WeightsColumn

Column name of a weights column

MaxMem

Set the maximum amount of memory you'd like to dedicate to the model run. E.g. "32G"

NThreads

Set the number of threads you want to dedicate to the model building

ModelID

A character string to name your model and output

ReturnModelObjects

Set to TRUE to output all modeling objects (E.g. plots and evaluation metrics)

model_path

A character string of your path file to where you want your output saved

metadata_path

A character string of your path file to where you want your model evaluation output saved. If left NULL, all output will be saved to model_path.

SaveModelObjects

Set to TRUE to return all modeling objects to your environment

SaveInfoToPDF

Set to TRUE to save insights to PDF

IfSaveModel

Set to "mojo" to save a mojo file, otherwise "standard" to save a regular H2O model object

H2OShutdown

Set to TRUE to shutdown H2O inside the function

H2OStartUp

Defaults to TRUE which means H2O will be started inside the function

TransformNumericColumns

Set to NULL to do nothing; otherwise supply the column names of numeric variables you want transformed

Methods

Choose from "YeoJohnson", "BoxCox", "Asinh", "Log", "LogPlus1", "Sqrt", "Asin", or "Logit". If more than one is selected, the one with the best normalization pearson statistic will be used. Identity is automatically selected and compared.

MaxModelsInGrid

Number of models to test from grid options (1080 total possible options)

NumOfParDepPlots

Tell the function the number of partial dependence calibration plots you want to create. Calibration boxplots will only be created for numerical features (not dummy variables)

GridTune

Set to TRUE to run a grid tuning procedure. Set a number in MaxModelsInGrid to tell the procedure how many models you want to test.

GridStrategy

"RandomDiscrete" or "Cartesian"

StoppingRounds

Iterations in grid tuning

MaxRunTimeSecs

Max run time in seconds

Distribution

"multinomial"

eval_metric

This is the metric used to identify best grid tuned model. Choose from "logloss"

RandomDistribution

Random effects family. Defaults NULL, otherwise it will run a hierarchical glm

RandomLink

Random effects link. Defaults NULL, otherwise it will run a hierarchical glm

Solver

Default "AUTO". Options include "IRLSM", "L_BFGS", "COORDINATE_DESCENT_NAIVE", "COORDINATE_DESCENT", "GRADIENT_DESCENT_LH", "GRADIENT_DESCENT_SQERR"

Alpha

Default 0.5 Otherwise supply a value between 0 and 1. 1 is equivalent to Lasso regression. 0 is equivalent to Ridge regression. Inbetween for a blend of the two.

Lambda

Default NULL. Regularization strength.

LambdaSearch

Default FALSE.

NLambdas

Default -1

Standardize

Default TRUE. Standardize numerical columns

RemoveCollinearColumns

Default FALSE. Removes some of the linearly dependent columns

InterceptInclude

Default TRUE

NonNegativeCoefficients

Default FALSE

link

"family_default"

Value

Saves to file and returned in list: VariableImportance.csv, Model, ValidationData.csv, EvaluationMetrics.csv, GridCollect, and GridList

Examples

Run this code

# NOT RUN {
# Create some dummy correlated data with numeric and categorical features
data <- RemixAutoML::FakeDataGenerator(
  Correlation = 0.85,
  N = 1000L,
  ID = 2L,
  ZIP = 0L,
  AddDate = FALSE,
  Classification = FALSE,
  MultiClass = TRUE)

# Run function
TestModel <- RemixAutoML::AutoH2oGLMMultiClass(

    # Compute management
    MaxMem = {gc();paste0(as.character(floor(as.numeric(system("awk '/MemFree/ {print $2}' /proc/meminfo", intern=TRUE)) / 1000000)),"G")},
    NThreads = max(1, parallel::detectCores()-2),
    H2OShutdown = TRUE,
    H2OStartUp = TRUE,
    IfSaveModel = "mojo",

    # Model evaluation:
    eval_metric = "logloss",
    NumOfParDepPlots = 3,

    # Metadata arguments:
    model_path = NULL,
    metadata_path = NULL,
    ModelID = "FirstModel",
    ReturnModelObjects = TRUE,
    SaveModelObjects = FALSE,
    SaveInfoToPDF = FALSE,

    # Data arguments:
    data = data,
    TrainOnFull = FALSE,
    ValidationData = NULL,
    TestData = NULL,
    TargetColumnName = "Adrian",
    FeatureColNames = names(data)[!names(data) %in% c("IDcol_1", "IDcol_2","Adrian")],
    RandomColNumbers = NULL,
    InteractionColNumbers = NULL,
    WeightsColumn = NULL,
    TransformNumericColumns = NULL,
    Methods = c("BoxCox", "Asinh", "Asin", "Log", "LogPlus1", "Sqrt", "Logit", "YeoJohnson"),

    # Model args
    GridTune = FALSE,
    GridStrategy = "Cartesian",
    StoppingRounds = 10,
    MaxRunTimeSecs = 3600 * 24 * 7,
    MaxModelsInGrid = 10,
    Distribution = "multinomial",
    Link = "family_default",
    RandomDistribution = NULL,
    RandomLink = NULL,
    Solver = "AUTO",
    Alpha = 0.5,
    Lambda = NULL,
    LambdaSearch = FALSE,
    NLambdas = -1,
    Standardize = TRUE,
    RemoveCollinearColumns = FALSE,
    InterceptInclude = TRUE,
    NonNegativeCoefficients = FALSE)
# }

Run the code above in your browser using DataLab

Description

Usage

Arguments

Value

See Also

Examples