categorical_encoding

  • Available in: GBM, DRF, Deep Learning, K-Means, Aggregator, XGBoost, Isolation Forest
  • Hyperparameter: yes

Description

This option specifies the encoding scheme to use for handling categorical features. Available schemes include the following:

GBM/DRF/Isolation Forest

  • auto or AUTO: Allow the algorithm to decide (default). For GBM, DRF, and Isolation Forest, the algorithm will perform Enum encoding when auto option is specified.
  • enum or Enum: Leave the dataset as is, internally map the strings to integers, and use these integers to make splits - either via ordinal nature when nbins_cats is too small to resolve all levels or via bitsets that do a perfect group split. Each category is a separate category; its name (or number) is irrelevant. For example, after the strings are mapped to integers for Enum, you can split {0, 1, 2, 3, 4, 5} as {0, 4, 5} and {1, 2, 3}.
  • enum_limited or EnumLimited: Automatically reduce categorical levels to the most prevalent ones during Aggregator training and only keep the T (1024) most frequent levels.
  • one_hot_explicit or OneHotExplicit: N+1 new columns for categorical features with N levels
  • binary or Binary: No more than 32 columns per categorical feature
  • eigen or Eigen: k columns per categorical feature, keeping projections of one-hot-encoded matrix onto k-dim eigen space only
  • label_encoder or LabelEncoder: Convert every enum into the integer of its index (for example, level 0 -> 0, level 1 -> 1, etc.) The categories are lexicographically mapped to numbers and lose their categorical nature, becoming ordinal. After the strings are mapped to integers, you can split {0, 1, 2, 3, 4, 5} as {0, 1, 2} and {3, 4, 5}.
  • sort_by_response or SortByResponse: Reorders the levels by the mean response (for example, the level with lowest response -> 0, the level with second-lowest response -> 1, etc.). This is useful in GBM/DRF, for example, when you have more levels than nbins_cats, and where the top level splits now have a chance at separating the data with a split. Note that this requires a specified response column.

Deep Learning

  • auto or AUTO: Allow the algorithm to decide. For Deep Learning, the algorithm will perform One Hot Internal encoding when auto is specified.
  • one_hot_internal or OneHotInternal: Leave the dataset as is. This internally expands each row via one-hot encoding on the fly. (default)
  • binary or Binary: No more than 32 columns per categorical feature
  • eigen or Eigen: k columns per categorical feature, keeping projections of one-hot-encoded matrix onto k-dim eigen space only
  • label_encoder or LabelEncoder: Convert every enum into the integer of its index (for example, level 0 -> 0, level 1 -> 1, etc.). The categories are lexicographically mapped to numbers and lose their categorical nature, becoming ordinal. After the strings are mapped to integers, you can split {0, 1, 2, 3, 4, 5} as {0, 1, 2} and {3, 4, 5}. This is useful for keeping the number of columns small for XGBoost or DeepLearning, where the algorithm otherwise perform ExplicitOneHotEncoding.
  • sort_by_response or SortByResponse: Reorders the levels by the mean response (for example, the level with lowest response -> 0, the level with second-lowest response -> 1, etc.). Note that this requires a specified response column.

Note: For Deep Learning, this value defaults to one_hot_internal. Similarly, if auto is specified, then the algorithm performs one_hot_internal encoding.

Aggregator

  • auto or AUTO: Allow the algorithm to decide. For Aggregator, the algorithm will perform One Hot Internal encoding when auto is specified.
  • one_hot_internal or OneHotInternal: Leave the dataset as is. This internally expands each row via one-hot encoding on the fly. (default)
  • binary or Binary: No more than 32 columns per categorical feature
  • eigen or Eigen: k columns per categorical feature, keeping projections of one-hot-encoded matrix onto k-dim eigen space only
  • label_encoder or LabelEncoder: Convert every enum into the integer of its index (for example, level 0 -> 0, level 1 -> 1, etc.). The categories are lexicographically mapped to numbers and lose their categorical nature, becoming ordinal. After the strings are mapped to integers, you can split {0, 1, 2, 3, 4, 5} as {0, 1, 2} and {3, 4, 5}. This is useful for keeping the number of columns small.
  • enum_limited or EnumLimited: Automatically reduce categorical levels to the most prevalent ones during Aggregator training and only keep the T (1024) most frequent levels.

XGBoost

  • auto or AUTO: Allow the algorithm to decide (default). In XGBoost, the algorithm will automatically perform one_hot_internal encoding. (default)
  • enum or Enum: 1 column per categorical feature. Each category is a separate category; its name (or number) is irrelevant. For example, after the strings are mapped to integers for Enum, you can split {0, 1, 2, 3, 4, 5} as {0, 4, 5} and {1, 2, 3}.
  • one_hot_internal or OneHotInternal: On the fly N+1 new cols for categorical features with N levels
  • one_hot_explicit or OneHotExplicit: N+1 new columns for categorical features with N levels
  • binary: No more than 32 columns per categorical feature
  • eigen or Eigen: k columns per categorical feature, keeping projections of one-hot-encoded matrix onto k-dim eigen space only
  • label_encoder or LabelEncoder: Convert every enum into the integer of its index (for example, level 0 -> 0, level 1 -> 1, etc.) The categories are lexicographically mapped to numbers and lose their categorical nature, becoming ordinal. After the strings are mapped to integers, you can split {0, 1, 2, 3, 4, 5} as {0, 1, 2} and {3, 4, 5}.
  • sort_by_response or SortByResponse: Reorders the levels by the mean response (for example, the level with lowest response -> 0, the level with second-lowest response -> 1, etc.). This is useful, for example, when you have more levels than nbins_cats, and where the top level splits now have a chance at separating the data with a split. Note that this requires a specified response column.
  • enum_limited or EnumLimited: Automatically reduce categorical levels to the most prevalent ones during training and only keep the T (1024) most frequent levels.

K-Means

  • auto or AUTO: Allow the algorithm to decide (default). For K-Means, the algorithm will perform Enum encoding when auto option is specified.
  • enum or Enum: Leave the dataset as is, internally map the strings to integers, and use these integers to make splits - either via ordinal nature when nbins_cats is too small to resolve all levels or via bitsets that do a perfect group split. Each category is a separate category; its name (or number) is irrelevant. For example, after the strings are mapped to integers for Enum, you can split {0, 1, 2, 3, 4, 5} as {0, 4, 5} and {1, 2, 3}.
  • one_hot_explicit or OneHotExplicit: N+1 new columns for categorical features with N levels
  • binary or Binary: No more than 32 columns per categorical feature
  • eigen or Eigen: k columns per categorical feature, keeping projections of one-hot-encoded matrix onto k-dim eigen space only
  • label_encoder or LabelEncoder: Convert every enum into the integer of its index (for example, level 0 -> 0, level 1 -> 1, etc.) The categories are lexicographically mapped to numbers and lose their categorical nature, becoming ordinal. After the strings are mapped to integers, you can split {0, 1, 2, 3, 4, 5} as {0, 1, 2} and {3, 4, 5}.

Example

library(h2o)
h2o.init()
# import the airlines dataset:
# This dataset is used to classify whether a flight will be delayed 'YES' or not "NO"
# original data can be found at http://www.transtats.bts.gov/
airlines <-  h2o.importFile("http://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")

# convert columns to factors
airlines["Year"] <- as.factor(airlines["Year"])
airlines["Month"] <- as.factor(airlines["Month"])
airlines["DayOfWeek"] <- as.factor(airlines["DayOfWeek"])
airlines["Cancelled"] <- as.factor(airlines["Cancelled"])
airlines['FlightNum'] <- as.factor(airlines['FlightNum'])

# set the predictor names and the response column name
predictors <- c("Origin", "Dest", "Year", "UniqueCarrier", "DayOfWeek", "Month", "Distance", "FlightNum")
response <- "IsDepDelayed"

# split into train and validation
airlines.splits <- h2o.splitFrame(data =  airlines, ratios = .8, seed = 1234)
train <- airlines.splits[[1]]
valid <- airlines.splits[[2]]

# try using the `categorical_encoding` parameter:
encoding = "OneHotExplicit"

# train your model
airlines_gbm <- h2o.gbm(x = predictors, y = response, training_frame = train, validation_frame = valid,
                        categorical_encoding = encoding, seed = 1234)

# print the auc for the validation set
print(h2o.auc(airlines_gbm, valid=TRUE))

import h2o
from h2o.estimators.gbm import H2OGradientBoostingEstimator
h2o.init()
h2o.cluster().show_status()

# import the airlines dataset:
# This dataset is used to classify whether a flight will be delayed 'YES' or not "NO"
# original data can be found at http://www.transtats.bts.gov/
airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")

# convert columns to factors
airlines["Year"]= airlines["Year"].asfactor()
airlines["Month"]= airlines["Month"].asfactor()
airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
airlines["Cancelled"] = airlines["Cancelled"].asfactor()
airlines['FlightNum'] = airlines['FlightNum'].asfactor()

# set the predictor names and the response column name
predictors = ["Origin", "Dest", "Year", "UniqueCarrier", "DayOfWeek", "Month", "Distance", "FlightNum"]
response = "IsDepDelayed"

# split into train and validation sets
train, valid= airlines.split_frame(ratios = [.8], seed = 1234)

# try using the `categorical_encoding` parameter:
encoding = "one_hot_explicit"

# initialize the estimator
airlines_gbm = H2OGradientBoostingEstimator(categorical_encoding = encoding, seed =1234)

# then train the model
airlines_gbm.train(x = predictors, y = response, training_frame = train, validation_frame = valid)

# print the auc for the validation set
airlines_gbm.auc(valid=True)