arules 0.0.0

multi-purpose association rules analysis

Arules - multi-purpose association rules

Arules is an open-source python package for association rules creation. It allows creation of association rules over tabular data (pandas dataframe). While standard association rules require transactional data, arules considers association rules as an analysis utility for categorical data. The Package also supports association rules over continuous data by application of binning methods (some basic methods are included in the package but users can define their own binning functions).

Installation

Python 3.6+ | Linux, Mac OS X, Windows

pip install -U arules

Getting Started

Let's create some association rules over some tabular data

import pandas as pd

anes96 = pd.read_csv("anes96.csv")
anes96.head()

| popul | TVnews | selfLR                 | ClinLR            | DoleLR                | PID              | age  | educ                 | income                   | vote    | logpopul           |
|-------|--------|------------------------|-------------------|-----------------------|------------------|------|----------------------|--------------------------|---------|--------------------|
| 0.0   | 7.0    | Extremely Conservative | Extremely liberal | Conservative          | Strong Republica | 36.0 | High school graduate | None or less than $2,999 | Dole    | -2.302585092994045 |
| 190.0 | 1.0    | Slightly liberal       | Slightly liberal  | Slightly conservative | Weak Democrat    | 20.0 | Some college         | None or less than $2,999 | Clinton | 5.247550249494384  |
| 31.0  | 7.0    | Liberal                | Liberal           | Conservative          | Weak Democrat    | 24.0 | Master's degree      | None or less than $2,999 | Clinton | 3.4372078191851885 |
| 83.0  | 4.0    | Slightly liberal       | Moderate          | Slightly conservative | Weak Democrat    | 28.0 | Master's degree      | None or less than $2,999 | Clinton | 4.4200447018614035 |
| 640.0 | 7.0    | Slightly conservative  | Conservative      | Moderate              | Strong Democrat  | 68.0 | Master's degree      | None or less than $2,999 | Clinton | 6.461624414147957  |

Note that the table contains both categorical and continuous fields (which can be handled using a selected binning method). Now we use arules to extract association rules according to a specification of interest

import arules as ar
from arules.utils import five_quantile_based_bins, top_bottom_10, top_5_variant_variables

rules, supp_dict = ar.create_association_rules(anes96,max_cols=2,binning_method=five_quantile_based_bins)

After the calculation is done we can present rules of selection for analysis purposes

ar.present_rules_per_consequent(rules,consequent={'vote':'Clinton'},
                                selection_function=top_5_variant_variables, drop_dups=True,
                                plot=True)

As we set the consequent to be: {'vote':'Clinton'}, the presented rules reflect the likelihood of an individual to vote for clinton given the respective feature. For example, if we consider the income variable above, a person with an income of 3,000-4,999 (which populates, according to the barchart, 1% of the sample) is approximately 1.6 times more likely (w.r.t. the average) to vote for Clinton, while a person with an income of 90,000-104,999 (which populates, according to the barchart, 4% of the sample) is approximately 1.4 times less likely to vote for Clinton.

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

Versioning

We use SemVer for versioning. For the versions available.

Authors

• Abir Koren - Initial work - WindWard

License

This project is licensed under the MIT License - see the LICENSE.md file for details