RuleFit

Introduction

H2O’s Rulefit algorithm combines tree ensembles and linear models to take advantage of both methods: the accuracy of a tree ensemble and the interpretability of a linear model.

The general algorithm fits a tree ensemble to the data, builds a rule ensemble by traversing each tree, evaluates the rules on the data to build a rule feature set, and fits a sparse linear model (LASSO) to the rule feature set joined with the original feature set.

MOJO Support

RuleFit supports importing and exporting MOJOs.

Defining a RuleFit Model (Beta API)

  • model_id: (Optional) Specify a custom name for the model to use as a reference. By default, H2O automatically generates a destination key.

  • training_frame: (Required) Specify the dataset used to build the model.

    Note: In Flow, if you click the Build a model button from the Parse cell, the training frame is entered automatically.

  • validation_frame: (Optional) Specify the dataset used to evaluate the accuracy of the model.

  • seed: Specify the random number generator (RNG) seed for algorithm components dependent on randomization. The seed is consistent for each H2O instance so that you can create models with the same starting conditions in alternate configurations. This value defaults to -1 (time-based random number).

  • y: (Required) Specify the column to use as the dependent variable.

    • For a regression model, this column must be numeric (Real or Int).

    • For a classification model, this column must be categorical (Enum or String). If the family is Binomial, the dataset cannot contain more than two levels.

  • x: Specify a vector containing the names or indicies of the predictor variables to use when building the model. If x is missing, then all columns except y are used.

  • algorithm: The algorithm to use to fit a tree ensemble. Must be one of: “AUTO”, “DRF”, or “GBM”. Defaults to “DRF”.

  • min_rule_length: Specify the minimal depth of trees to be fit. Defaults to 3.

  • max_rule_length: Specify the maximal depth of trees to be fit. Defaults to 3.

  • max_num_rules: The maximum number of rules to return. Defaults to -1, which means the number of rules are selected by diminishing returns in model deviance.

  • model_type: Specify the type of base learners in the ensemble. Must be one of: “rules_and_linear”, “rules”, or “linear”. Defaults to “rules_and_linear”.

    • If the model_type is rules_and_linear, the algorithm fits a linear model to the rule feature set joined with the original feature set.

    • If the model_type is rules, the algorithm fits a linear model only to the rule feature set (no linear terms can become important).

    • If the model_type is linear, the algorithm fits a linear model only to the original feature set (no rule terms can become important).

  • rule_generation_ntrees: Specify the number of trees for tree ensemble. Defaults to 50.

  • weights_column: Specify a column to use for the observation weights, which are used for bias correction. The specified weights_column must be included in the specified training_frame.

    Python only: To use a weights column when passing an H2OFrame to x instead of a list of column names, the specified training_frame must contain the specified weights_column.

    Note: Weights are per-row observation weights and do not increase the size of the data frame. This is typically the number of times a row is repeated, but non-integer values are supported as well. During training, rows with higher weights matter more due to the larger loss function pre-factor.

  • distribution: Specify the distribution (i.e. the loss function). The options are AUTO, bernoulli, multinomial, gaussian, poisson, gamma, laplace, quantile, huber, or tweedie.

    • If the distribution is bernoulli, the response column must be 2-class categorical.

    • If the distribution is quasibinomial, the response column must be numeric and binary.

    • If the distribution is multinomial, the response column must be categorical.

    • If the distribution is poisson, the response column must be numeric.

    • If the distribution is tweedie, the response column must be numeric.

    • If the distribution is gaussian, the response column must be numberic.

    • If the distribution is gamma, the response column must be numeric.

    • If the distribution is fractionalbinomial, the response column must be numeric between 0 and 1.

    • If the distribution is negativebinomial, the response column must be numeric and non-negative.

    • If the distribution is ordinal, the response column must be categorical with at least 3 levels.

    • If the distribution is AUTO,

      • and the response is Enum with cardinality = 2, then the family is automatically determined as bernoulli.

      • and the response is Enum with cardinality > 2, then the family is automatically determined as multinomial.

      • and the response is numeric (Real or Int), then the family is automatically determined as gaussian.

  • remove_duplicates: Specify whether to remove rules which are identical to an earlier rule. Defaults to true.

  • lambda: Specify the regularization strength for LASSO regressor.

  • max_categorical_levels: Rulefit handles categorical features by EnumLimited scheme. That means it automatically reduces categorical levels to the most prevalent ones and only keeps the max_categorical_levels most frequent levels. Defaults to 10.

Interpreting a RuleFit Model

The output for the RuleFit model includes:

  • model parameters

  • rule importances in tabular form

  • training and validation metrics of the underlying linear model

Examples

R

Python

library(h2o)
h2o.init()

# Import the titanic dataset:
f <- "https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv"
coltypes <- list(by.col.name = c("pclass", "survived"), types=c("Enum", "Enum"))
df <- h2o.importFile(f, col.types = coltypes)

# Split the dataset into train and test
splits <- h2o.splitFrame(data = df, ratios = 0.8, seed = 1)
train <- splits[[1]]
test <- splits[[2]]

# Set the predictors and response; set the factors:
response <- "survived"
predictors <- c("age", "sibsp", "parch", "fare", "sex", "pclass")

# Build and train the model:
rfit <- h2o.rulefit(y = response,
                    x = predictors,
                    training_frame = train,
                    max_rule_length = 10,
                    max_num_rules = 100,
                    seed = 1)

# Retrieve the rule importance:
print(rfit@model$rule_importance)

# Predict on the test data:
h2o.predict(rfit, newdata = test)

import h2o
h2o.init()
from h2o.estimators import H2ORuleFitEstimator

# Import the titanic dataset and set the column types:
f = "https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv"
df = h2o.import_file(path=f, col_types={'pclass': "enum", 'survived': "enum"})

# Split the dataset into train and test
train, test = df.split_frame(ratios=[0.8], seed=1)

# Set the predictors and response:
x = ["age", "sibsp", "parch", "fare", "sex", "pclass"]
y = "survived"

# Build and train the model:
rfit = H2ORuleFitEstimator(max_rule_length=10,
                           max_num_rules=100,
                           seed=1)
rfit.train(training_frame=train, x=x, y=y)

# Retrieve the rule importance:
print(rfit.rule_importance())

# Predict on the test data:
rfit.predict(test)