FPGrowth
Description
The FPGrowth (Frequent Pattern Growth) function uses an
FP-growth algorithm to generate association rules from patterns
in a data set, and then determines their interestingness.
Usage
td_fpgrowth_mle (
data = NULL,
tran.item.columns = NULL,
tran.id.columns = NULL,
patterns.or.rules = "both",
groupby.columns = NULL,
pattern.distribution.key.column = NULL,
rule.distribution.key.column = NULL,
compress = "nocompress",
group.size = 4,
min.support = 0.05,
min.confidence = 0.8,
max.pattern.length = "2",
antecedent.count.range = "1-INFINITE",
consequence.count.range = "1-1",
delimiter = ",",
data.sequence.column = NULL
)
Arguments
data |
Required Argument.
Specifies the name of the tbl_teradata that contains the data set.
|
tran.item.columns |
Required Argument.
Specifies the names of the columns that contain transaction items to
analyze.
Types: character OR vector of Strings (character)
|
tran.id.columns |
Required Argument.
Specifies the names of the columns that contain identifiers for the
transaction items.
Types: character OR vector of Strings (character)
|
patterns.or.rules |
Optional Argument.
Specifies whether the function outputs patterns, rules, or both.
An example of a pattern is onions, potatoes, hamburger.
Default Value: "both"
Permitted Values: both, patterns, rules
Types: character
|
groupby.columns |
Optional Argument.
Specifies the names of columns that define the partitions into which
the function groups the input data and calculates output for it. At
least one column must be usable as a distribution key. If you omit
this argument, then the function considers all input data to be in a
single partition.
Note: Do not specify the same column in both this
argument and the "tran.id.column argument", because this causes
incorrect counting in the partitions.
Types: character OR vector of Strings (character)
|
pattern.distribution.key.column |
Optional Argument.
Specifies the name of the column to use as the distribution key for
"output.pattern.table" tbl_teradata. The default value is the first
column name - "pattern_<tran.item.columns>" as generated in the
"output.pattern.table" tbl_teradata.
Note: Only one column name can be specified.
Types: character
|
rule.distribution.key.column |
Optional Argument.
Specifies the name of the column to use as the distribution key for
"output.rule.table" tbl_teradata. The default value is the first
column name - "antecedent_<tran.item.columns>" as generated
in the "output.rule.table" tbl_teradata.
Note: Only one column name can be specified.
Types: character
|
compress |
Optional Argument.
Specifies the compression level of the output tbl_teradata objects.
Realized compression ratios depend on both this value and the data
characteristics. These ratios typically range from 3x to 12x.
Default Value: "nocompress"
Permitted Values: low, medium, high, nocompress
Types: character
|
group.size |
Optional Argument.
Specifies the number of transaction items to be assigned to each
worker. This value must be an integer in the range from 1 to the
number of distinct transaction items, inclusive. For a machine with
limited RAM, use a relatively small value.
Default Value: 4
Types: integer
|
min.support |
Optional Argument.
Specifies the minimum support value of returned patterns (including
the specified support value). This value must be a floating point number in the
range [0, 1].
Default Value: 0.05
Types: numeric
|
min.confidence |
Optional Argument.
Specifies the minimum confidence value of returned patterns
(including the specified confidence value). This value must be a
floating point number in the range [0, 1].
Default Value: 0.8
Types: numeric
|
max.pattern.length |
Optional Argument.
Specifies the maximum length of returned patterns. The length of a
pattern is the sum of the item numbers in the antecedent and
consequence columns. This value must be an integer greater than 2.
"max.pattern.length" also limits the length of
returned rules to this value.
Default Value: "2"
Types: character
|
antecedent.count.range |
Optional Argument.
Specifies the range for the number of items in the antecedent.
This argument takes the format, "lower_bound-upper_bound".
The function returns only patterns for which number of items in the
antecedent is in the range [lower_bound, upper_bound].
The lower_bound must be an integer greater than 0.
The lower_bound and upper_bound can be equal.
Default Value: "1-INFINITE"
Types: character
|
consequence.count.range |
Optional Argument.
Specifies the range for the number of items in the consequence.
The function returns only patterns for which number of items in the consequence
is in the range [lower_bound, upper_bound]. The lower_bound must be greater an
integer greater than 0. The lower_bound and upper_bound can be equal.
Default Value: "1-1"
Types: character
|
delimiter |
Optional Argument.
Specifies the delimiter that separates items in the output.
Default Value: ","
Types: character
|
data.sequence.column |
Optional Argument.
Specifies the vector of column(s) that uniquely identifies each row
of the input argument "data". The argument is used to ensure
deterministic results for functions which produce results that vary
from run to run.
Types: character OR vector of Strings (character)
|
Value
Function returns an object of class "td_fpgrowth_mle" which is a
named list containing objects of class "tbl_teradata".
Named list members can be referenced directly with the "$" operator
using the following names:
output.pattern.table
-
output.rule.table
output
Examples
# Get the current context/connection
con <- td_get_context()$connection
# Load example data.
loadExampleData("fpgrowth_example", "sales_transaction")
# Create object(s) of class "tbl_teradata".
# Sales transaction data of an office supply chain store.
# The column "product" specifies the items that are purchased by a
# customer in a given transaction (column "orderid")
sales_transaction <- tbl(con, "sales_transaction")
# Compute association rules based on the pattern in the "product" column
td_fpgrowth_out <- td_fpgrowth_mle(data = sales_transaction,
tran.item.columns = "product",
tran.id.columns = "orderid",
groupby.columns = "region",
min.support = 0.01,
min.confidence = 0.0,
max.pattern.length = "4"
)