Frequent Itemset Mining

This repository provides C++ implementations of popular frequent itemset mining algorithms: Apriori, FP-Growth, ECLAT, and RElim. The goal of this project is to provide solutions for mining frequent patterns from datasets in the context of association rule learning.

The repository includes a suite of Google Unit Tests, a benchmarking tool to evaluate algorithm performance, and a command-line interface (CLI) for easy usage.

1. Introduction

Frequent itemset mining is a crucial problem in data mining, used to discover associations between items in large datasets. This repository offers implementations of four well-known algorithms, each with its own strengths and weaknesses.

2. Algorithms

2.1 Apriori Algorithm

The Apriori algorithm is one of the most popular algorithms for mining frequent itemsets. It uses a level-wise search strategy where it generates candidate itemsets of length $k$ from itemsets of length $k-1$, pruning the search space based on the minimum support threshold.

2.2 FP-Growth Algorithm

FP-Growth (Frequent Pattern Growth) is a more efficient algorithm compared to Apriori. It uses a compact tree structure (FP-tree) to represent the dataset and avoid the need for candidate generation.

2.3 ECLAT Algorithm

ECLAT (Equivalence Class Transformation) is another algorithm used for frequent itemset mining. It works by converting the problem into set intersections, efficiently finding frequent itemsets through vertical data representation.

2.4 RElim Algorithm

RElim (Recursive Elimination of Minimum Support) is an algorithm for frequent itemset mining. It recursively eliminates itemsets that do not meet the minimum support threshold, reducing the search space and improving efficiency. Unlike the FP-Growth algorithm, RElim uses a simpler data structure.

3. Command-Line Interface (CLI)

The repository includes a simple CLI for interacting with the algorithms and running tests. The following table shows the available commands:

Command	Description
--help	Show a help text.
--override	If set, the output file is overwritten if it already exists.
-i, --input	Path to the input file containing the database.
-o, --output	Path to the output file where the frequent itemsets will be saved.
-s, --min-support	Minimum support threshold for the frequent itemsets.
-a, --algorithm	Specifies which algorithm to use. Valid values are apriori, fp-growth, eclat, relim.

Example usage:

fim --input data_01.csv --output result_relim_01.csv --min-support 0.6 --algorithm relim --override

4. Conclusion

This repository provides implementations of popular frequent item set mining algorithms in C++. You can test the algorithms on your own datasets using the CLI and measure the performance using the benchmarking tools.

We hope that this project will be useful for both learning purposes and practical applications in data mining.

Author

Roland Abel

License

This software is available under the following licenses:

See MIT License (MIT)