discretizer_fn: Discretizer function

In robertzk/syberiaMungebits: Common mungebits used with Syberia projects.

Description

Discretizer function

Usage

discretizer_fn(column, granularity = 3, mode_freq_threshold = 0.15,
  mode_ratio_threshold = 1.5, category_range = min(granularity, 20):20,
  lower_count_bound = granularity, upper_count_bound = NULL,
  missing_level = "Missing", ...)

Arguments

column	an atomic vector. The variable to discretize.
granularity	an integer. The suggested number of levels.
mode_freq_threshold	a real value between 0 and 1. If the mode of the variable exceeds this value and is greater than mode_ratio_threshold (see next parameter) times the next greatest mode (i.e., the ratio of the value occuring most often over the value occuring second most often is over mode_ratio_threshold) then the variable will be attempted to be discretized in manner as to make the mode its own bucket. (so if the mode is 5, we'd want, e.g., [2,4), 5, and (5, 7]).
mode_ratio_threshold	a real value. See the mode_freq_threshold parameter.
category_range	The number of levels to consider when the discretization procedure descrized in the mode_freq_threshold parameter is employed. The default is min(granularity, 20):20.
lower_count_bound	an integer. Variables with less than or equal to this many unique values will not get discretized. Default is granularity.
upper_count_bound	an integer. Variables with more than or equal to this many unique values will not get discretized. Default is granularity.
missing_level	character. Any values that were NA prior to discretization will be replaced with this level. If set to NULL, then the NAs will remain. The default is "Missing".
...	additional arguments to pass to arules::discretize.

robertzk/syberiaMungebits documentation built on July 30, 2019, 3:37 p.m.