🚨 Repository Moved! 🚨

Important Notice: This repository has been moved to a new location. All future updates, issues, and contributions will be handled at our new repository:

🔗 https://github.com/lukassykora/action-rules

Please update your bookmarks and direct any questions, pull requests, or issues to the new repository.

Thank you for your continued support and contributions!

Action Rules

Action Rules (actionrules) is an implementation of Action Rules from Classification Rules algorithm described in

Dardzinska, A. (2013). Action rules mining. Berlin: Springer.

If you use this package, please cite:

Sýkora, Lukáš, and Tomáš Kliegr. "Action Rules: Counterfactual Explanations in Python." RuleML Challenge 2020. CEUR-WS. http://ceur-ws.org/Vol-2644/paper36.pdf

GIT repository

https://github.com/lukassykora/actionrules

Installation

pip install actionrules-lukassykora

Jupyter Notebooks

• Titanic It is the best explanation of all possibilities.
• Telco A brief demonstration.
• Ras Based on the example in (Ras, Zbigniew W and Wyrzykowska, ARAS: Action rules discovery based on agglomerative strategy, 2007).
• Attrition High-Utility Action Rules Mining example.

Example 1

Get data from csv. Get action rules from classification rules. Classification rules have confidence 55% and support 3%. Stable part of action rule is "Age". Flexible attributes are "Embarked", "Fare", "Pclass". Target is a Survived value 1.0. No nan values. Use reduction tables for speeding up. Minimal 1 stable antecedent Minimal 1 flexible antecedent

from actionrules.actionRulesDiscovery import ActionRulesDiscovery

actionRulesDiscovery = ActionRulesDiscovery()
actionRulesDiscovery.read_csv("data/titanic.csv", sep="\t")
actionRulesDiscovery.fit(stable_attributes = ["Age"],
                         flexible_attributes = ["Embarked", "Fare", "Pclass"],
                         consequent = "Survived",
                         conf=55,
                         supp=3,
                         desired_classes = ["1.0"],
                         is_nan=False,
                         is_reduction=True,
                         min_stable_attributes=1,
                         min_flexible_attributes=1,
                         max_stable_attributes=5,
                         max_flexible_attributes=5)
actionRulesDiscovery.get_action_rules()

The output is a list where the first part is an action rule and the second part is a tuple of (support before, support after, action rule support) and (confidence before, confidence after, action rule confidence).

Example 2

Get data from pandas dataframe. Get action rules from classification rules. Classification rules have confidence 50% and support 3%. Stable attributes are "Age" and "Sex". Flexible attributes are "Embarked", "Fare", "Pclass". Target is a Survived that changes from 0.0 to 1.0. No nan values. Use reduction tables for speeding up. Minimal 1 stable antecedent Minimal 1 flexible antecedent

from actionrules.actionRulesDiscovery import ActionRulesDiscovery
import pandas as pd

dataFrame = pd.read_csv("data/titanic.csv", sep="\t")
actionRulesDiscovery = ActionRulesDiscovery()
actionRulesDiscovery.load_pandas(dataFrame)
actionRulesDiscovery.fit(stable_attributes = ["Age", "Sex"],
                         flexible_attributes = ["Embarked", "Fare", "Pclass"],
                         consequent = "Survived",
                         conf=50,
                         supp=3,
                         desired_changes = [["0.0", "1.0"]],
                         is_nan=False,
                         is_reduction=True,
                         min_stable_attributes=1,
                         min_flexible_attributes=1,
                         max_stable_attributes=5,
                         max_flexible_attributes=5)
actionRulesDiscovery.get_pretty_action_rules()

The output is a list of action rules in pretty text form.