GSP-Py Documentation

Welcome to the official documentation for GSP-Py, a Python library that implements the Generalized Sequential Pattern (GSP) algorithm. Use this site to install the library, explore the CLI, and dive into accelerated backends.

Highlights

• Sequence mining with support-based pruning and candidate generation.
• Sequence abstraction for typed pattern representation with rich metadata support.
• Multiple data formats including JSON, CSV, SPM/GSP, Parquet, and Arrow.
• Token mapping utilities for transparent string ↔ integer conversion.
• Optional acceleration via Rust and GPU backends.
• Developer-friendly tools including typed APIs and optional benchmarks.

Quickstart

Install the package from PyPI and run a simple search:

pip install gsppy

from gsppy.gsp import GSP

transactions = [
    ["Bread", "Milk"],
    ["Bread", "Diaper", "Beer", "Eggs"],
    ["Milk", "Diaper", "Beer", "Coke"],
]

patterns = GSP(transactions).search(min_support=0.3)
for level, freq_patterns in enumerate(patterns, start=1):
    print(f"Level {level}: {freq_patterns}")

Using Sequence Objects (New in 4.0+)

Get richer pattern representation with typed Sequence objects:

from gsppy import GSP

transactions = [
    ["Bread", "Milk"],
    ["Bread", "Diaper", "Beer", "Eggs"],
    ["Milk", "Diaper", "Beer", "Coke"],
]

# Enable Sequence objects
patterns = GSP(transactions).search(min_support=0.3, return_sequences=True)

for level_patterns in patterns:
    for seq in level_patterns:
        print(f"{seq.items} (support={seq.support}, length={seq.length})")
        if "Milk" in seq:
            enriched = seq.with_metadata(confidence=0.85)

Loading from SPM/GSP Format

from gsppy.utils import read_transactions_from_spm
from gsppy import GSP

# Load classical SPM format with -1/-2 delimiters
transactions = read_transactions_from_spm('data.txt')
patterns = GSP(transactions).search(min_support=0.3)

Continue to the Usage guide for installation details and backend selection tips.