arm-preprocessing 0.2.0

Implementation of several preprocessing techniques for Association Rule Mining (ARM)

arm-preprocessing

• Free software: MIT license
• Documentation: http://arm-preprocessing.readthedocs.io
• Python: 3.9.x, 3.10.x, 3.11.x, 3.12x
• Tested OS: Windows, Ubuntu, Fedora, Alpine, Arch, macOS. However, that does not mean it does not work on others

About 📋

arm-preprocessing is a lightweight Python library supporting several key steps involving data preparation, manipulation, and discretisation for Association Rule Mining (ARM). 🧠 Embrace its minimalistic design that prioritises simplicity. 💡 The framework is intended to be fully extensible and offers seamless integration with related ARM libraries (e.g., NiaARM). 🔗

Why arm-preprocessing?

While numerous libraries facilitate data mining preprocessing tasks, this library is designed to integrate seamlessly with association rule mining. It harmonises well with the NiaARM library, a robust numerical association rule mining framework. The primary aim is to bridge the gap between preprocessing and rule mining, simplifying the workflow/pipeline. Additionally, its design allows for the effortless incorporation of new preprocessing methods and fast benchmarking.

Key features ✨

• Loading various formats of datasets (CSV, JSON, TXT) 📊
• Converting datasets to different formats 🔄
• Loading different types of datasets (numerical dataset, discrete dataset, time-series data, text, etc.) 📉
• Dataset identification (which type of dataset) 🔍
• Dataset statistics 📈
• Discretisation methods 📏
• Data squashing methods 🤏
• Feature scaling methods ⚖️
• Feature selection methods 🎯

Installation 📦

pip

To install arm-preprocessing with pip, use:

pip install arm-preprocessing

To install arm-preprocessing on Alpine Linux, please use:

$ apk add py3-arm-preprocessing

To install arm-preprocessing on Arch Linux, please use an AUR helper:

$ yay -Syyu python-arm-preprocessing

Usage 🚀

Data loading

The following example demonstrates how to load a dataset from a file (csv, json, txt). More examples can be found in the examples/data_loading directory:

• Loading a dataset from a CSV file
• Loading a dataset from a JSON file
• Loading a time-series dataset

from arm_preprocessing.dataset import Dataset

# Initialise dataset with filename (without format) and format (csv, json, txt)
dataset = Dataset('path/to/datasets', format='csv')

# Load dataset
dataset.load_data()
df = dataset.data

Missing values

The following example demonstrates how to handle missing values in a dataset using imputation. More examples can be found in the examples/missing_values directory:

• Handling missing values in a dataset using row deletion
• Handling missing values in a dataset using column deletion
• Handling missing values in a dataset using imputation

from arm_preprocessing.dataset import Dataset

# Initialise dataset with filename and format
dataset = Dataset('examples/missing_values/data', format='csv')
dataset.load()

# Impute missing data
dataset.missing_values(method='impute')

Data discretisation

The following example demonstrates how to discretise a dataset using the equal width method. More examples can be found in the examples/discretisation directory:

• Discretising a dataset using the equal width method
• Discretising a dataset using the equal frequency method
• Discretising a dataset using k-means clustering

from arm_preprocessing.dataset import Dataset

# Initialise dataset with filename (without format) and format (csv, json, txt)
dataset = Dataset('datasets/sportydatagen', format='csv')
dataset.load_data()

# Discretise dataset using equal width discretisation
dataset.discretise(method='equal_width', num_bins=5, columns=['calories'])

Data squashing

The following example demonstrates how to squash a dataset using the euclidean similarity. More examples can be found in the examples/squashing directory:

• Squashing a dataset using the euclidean similarity
• Squashing a dataset using the cosine similarity

from arm_preprocessing.dataset import Dataset

# Initialise dataset with filename and format
dataset = Dataset('datasets/breast', format='csv')
dataset.load()

# Squash dataset
dataset.squash(threshold=0.75, similarity='euclidean')

Feature scaling

The following example demonstrates how to scale the dataset's features. More examples can be found in the examples/scaling directory:

• Scale features using normalisation
• Scale features using standardisation

from arm_preprocessing.dataset import Dataset

# Initialise dataset with filename and format
dataset = Dataset('datasets/Abalone', format='csv')
dataset.load()

# Scale dataset using normalisation
dataset.scale(method='normalisation')

Feature selection

The following example demonstrates how to select features from a dataset. More examples can be found in the examples/feature_selection directory:

• Select features using the Kendall Tau correlation coefficient

from arm_preprocessing.dataset import Dataset

# Initialise dataset with filename and format
dataset = Dataset('datasets/sportydatagen', format='csv')
dataset.load()

# Feature selection
dataset.feature_selection(
    method='kendall', threshold=0.15, class_column='calories')

Related frameworks 🔗

[1] NiaARM: A minimalistic framework for Numerical Association Rule Mining

[2] uARMSolver: universal Association Rule Mining Solver

References 📚

[1] I. Fister, I. Fister Jr., D. Novak and D. Verber, Data squashing as preprocessing in association rule mining, 2022 IEEE Symposium Series on Computational Intelligence (SSCI), Singapore, Singapore, 2022, pp. 1720-1725, doi: 10.1109/SSCI51031.2022.10022240.

[2] I. Fister Jr., I. Fister A brief overview of swarm intelligence-based algorithms for numerical association rule mining. arXiv preprint arXiv:2010.15524 (2020).

License

This package is distributed under the MIT License. This license can be found online at http://www.opensource.org/licenses/MIT.

Disclaimer

This framework is provided as-is, and there are no guarantees that it fits your purposes or that it is bug-free. Use it at your own risk!