fpgrowth

This document has been machine translated.

Execute correlation pattern extraction by FP-Growth.

The execution method uses JSON-RPC v2.0.

Example request

fpgrowth is a kind of analysis API, and can be executed by specifying api_method="fpgrowth" in the process method of the Provenance API. The following is an example of starting a Provenance session, executing fpgrowth, and then exiting the session.

import xdata_prov.client import Api
api = Api()
api.begin_session()

api.process(api_method="fpgrowth", api_params={
    "output_ddc": "ddc:jartic_xrain_rules",
    "input_ddc": "ddc:jartic_xrain_items",
    "min_support": 0.01,
    "min_confidence": 0.75,
    "param_json": "{ (see parameter entry) }"
})

api.commit()
api.end_session()

Parameters.

When calling the process method with api_method="fpgrowth", api_params will take a dict containing the following keys. Parameters with a blank default value are required.

key	description	default value
output_ddc	output destination for processing results ddc
output_mode	output mode (overwrite or error)	error
input_ddc	input data ddc
min_support	minimum value of support
min_confidence	minimum value of confidence
param_json	Parameters passed to FP-Growth

filters.

You can specify the criteria for narrowing down the correlation patterns to be output, with a JSON array as the value. An example is shown below.

"filters": [
  {
    "pre": {"must": ["mesh", "rf"]},
    "post": {"must": ["cl"], "must_not": ["mesh", "dow", "peak"] }
  }
]

Each element of the array describes a precondition (pre) and a filtering condition for the consequent (post).

key	description
pre.must	list of categories that must be present in the presupposition (forward matching, must include all)
pre.must_not	list of categories that should not be in the premise clause (forward match, none of them)
post.must	list of categories that must be present in the consequent clause (forward matching, must include all)
post.must_not	list of categories that should not be present in the consequent clause (forward match, must not include any of them)

If any of the conditions listed as the value of filters are met, the output is targeted.

Input data

symbol-translated transaction table

The input ddc specified by input_ddc is a "symbol-transformed transaction table". This table must have the following schema.

column name	data type	description
id	integer	ID that uniquely identifies the transaction
start_datetime	timestamp with time zone	start date and time
end_datetime	timestamp with time zone	end_date
location	geometry	spatial range
meshcode	character varying	meshcode
items	text[]	a set of events that occurred in this space-time range

Output data

Correlation rule table

The "correlation rule table" will be output to the destination ddc specified by output_ddc. This table will have the following schema.

column name	data type	description
id	integer	ID that uniquely identifies the rule
pre	text[]	antecedent
post	text[]	consequent (conclusion)
support	integer	absolute support (the number of records satisfying the rule)
confidence	double precision	confidence (degree of confidence)

• support column in the rule table stores the number of records, rather than the degree of support

return value

fpgrowth returns the ddc information of the output destination ddc. This is the behavior defined by the specification of the process method of the Provenance API.