TemporalGeneralizedRules 1.0.2

Temporal Generalized Association Rules

This library provides four algorithms related to Association Rule mining. You can download this repository as a package with:

pip install TemporalGeneralizedRules

The algorithms are:

• Apriori
• Cumulate
• HTAR
• HTGAR

These algorithms use a transactional dataset that is transformed to a vertical format for optimization. Dataset MUST follow the following format:

order_id	product_name
1	Bread
1	Milk
2	Bread
2	Beer
3	Eggs

Or if timestamps are provided:

order_id	timestamp	product_name
1	852087600	Bread
1	852087600	Milk
2	854420400	Bread
2	854420400	Beer
3	854420400	Eggs

For taxonomy file use the following format (don't provide headers):

child	parent
Bread	Dairy
Milk	Dairy
Beer	Beverage

One line for each child, parent

Each field is separated by ","

TGAR

This is the main class that must be instantiated once.

Usage

import TemporalGeneralizedRules

tgar = TemporalGeneralizedRules.TGAR()

Apriori

This algorithm has four parameters:

• filepath: Filepath of the dataset in csv format with the format discussed in the previous section.
• min_supp: Minimum support threshold.
• min_conf: Minimum confidence threshold.
• parallel_count: Optional parameter that enables parallelization in candidate count phase of the algorithm.

Usage

tgar.apriori("dataset.csv", 0.05, 0.5)

Cumulate

This algorithm has six parameters:

• filepath: Filepath of the dataset in csv format with the format discussed in the previous section.
• taxonomy_filepath: Filepath of the taxonomy in csv format with the format discussed in the previous section.
• min_supp: Minimum support threshold.
• min_conf: Minimum confidence threshold.
• min_r: Minimum R-interesting threshold.
• parallel_count: Optional parameter that enables parallelization in candidate count phase of the algorithm. It can make execution faster.

Usage

tgar.cumulate("dataset.csv", 0.05, 0.5, 1.1)

HTAR

This algorithm has four parameters:

• filepath: Filepath of the dataset in csv format with the format discussed in the previous section.
• min_supp: Minimum support threshold.
• min_conf: Minimum confidence threshold.
• parallel_count: Optional parameter that enables parallelization in candidate count phase of the algorithm. It can make execution faster.

Usage

tgar.htar("dataset.csv", 0.05, 0.5)

HTGAR

This algorithm has six parameters:

• filepath: Filepath of the dataset in csv format with the format discussed in the previous section.
• taxonomy_filepath: Filepath of the taxonomy in csv format with the format discussed in the previous section.
• min_supp: Minimum support threshold.
• min_conf: Minimum confidence threshold.
• min_r: Minimum R-interesting threshold.
• parallel_count: Optional parameter that enables parallelization in candidate count phase of the algorithm. It can make execution faster.

Usage

tgar.htgar("dataset.csv", 0.05, 0.5, 1.1)

Pypy

For a better performance we recommend using this package with Pypy, a faster implementation of python.

https://www.pypy.org/

Bibliography

The algorithms provided in this library were based on the following papers:

• Rakesh Agrawal and Ramakrishnan Srikant. 1994. Fast Algorithms for Mining Association Rules in Large Databases. In Proceedings of the 20th International Conference on Very Large Data Bases (VLDB '94). Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 487–499. https://dl.acm.org/doi/10.5555/645920.672836

• Ramakrishnan Srikant, Rakesh Agrawal, Mining generalized association rules, Future Generation Computer Systems, Volume 13, Issues 2–3, 1997, Pages 161-180, ISSN 0167-739X. https://www.sciencedirect.com/science/article/pii/S0167739X97000198

• R. Agrawal and J. C. Shafer, "Parallel mining of association rules," in IEEE Transactions on Knowledge and Data Engineering, vol. 8, no. 6, pp. 962-969, Dec. 1996, doi: 10.1109/69.553164. https://ieeexplore.ieee.org/document/553164

• Hong et al., 2016.Hong, T.-P., Lan, G.-C., Su, J.-H., Wu, P.-S., and Wang, S.-L. (2016). Discovery of temporal association rules with hierarchical granular framework. Applied Computing and Informatics, 12(2):134–141 https://www.sciencedirect.com/science/article/pii/S2210832716000041