SCAMPI - Fast Pattern Discovery in Massive Time Series

This repository provides the reference implementation and experimental material for the paper “SCAMPI - Fast Pattern Discovery in Massive Time Series”.

Repository Structure

This repository contains the full framework, benchmark datasets, and reproducible experiments used in the evaluation.

• scampi/
  Core implementation of SCAMPI and the k-Motiflets algorithm.

• benchmarks/
  Benchmark scripts and helpers. Command-line benchmark callers live in benchmarks/cli/, conversion scripts in benchmarks/converters/, reusable benchmark runners in benchmarks/runners/, and plotting scripts in benchmarks/plots/.

• notebooks/
  Jupyter notebooks demonstrating typical use cases and reproducing paper figures.

• datasets/momp/
  large-Scale Benchmark time series datasets used throughout the paper.
  Please see https://sites.google.com/view/momp2024, too

• tests/
  Unit tests for the implementation.

• tests/results/
  Benchmark output CSVs and converted result files.

SCAMPI (SCalable Anytime Mining of Patterns In time series)

This paper introduces SCAMPI.

Installation

The easiest is to use pip to install motiflets.

a) Install using pip

BLINDED

You can also install the project from source.

b) Build from Source

Clone the repository:

git clone https://github.com/patrickzib/scampi.git
cd scampi

Install the package:

pip install .

Usage Example

from scampi.scampi import *

ml = SCAMPI(
    ds_name,  # dataset name
    series,   # time series data
    n_jobs,    # number of CPU cores
    scampi_max_memory = "1 GB",  # max-memory to use for SCAMPI
    scampi_delta = "0.1" 
)

k_max = 20  # maximum motif set size to consider
motif_length = 100  # length of the motifs to search for
top_N = 10  # number of top motif sets to return

dists, candidates, elbow_points = ml.fit_k_elbow(
    k_max,
    motif_length,
    top_N=top_N
)

ml.plot_motifset()

Benchmark Scripts

The MOMP benchmark entry points are:

python benchmarks/cli/run_momp_benchmarks.py --list-methods
python benchmarks/cli/run_momp_benchmarks.py --methods faiss-hnsw