Awesome

affiliation-metrics-py

Python 3 implementation of the affiliation metrics and tests for reproducing the experiments described in Local Evaluation of Time Series Anomaly Detection Algorithms, accepted in KDD 2022 Research Track: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.

Installation

Type pip install . to install the affiliation package. Only the standard Python library is needed, there is no dependency to external libraries.

Usage

In a Python session, the following lines give an example for computing the affiliation metrics from prediction and ground truth vectors:

from affiliation.generics import convert_vector_to_events
from affiliation.metrics import pr_from_events

vector_pred = [0, 0, 0, 0, 1, 0, 0, 0, 1, 0]
vector_gt   = [0, 0, 0, 1, 0, 0, 0, 1, 1, 1]

events_pred = convert_vector_to_events(vector_pred) # [(4, 5), (8, 9)]
events_gt = convert_vector_to_events(vector_gt)     # [(3, 4), (7, 10)]
Trange = (0, len(vector_pred))

pr_from_events(events_pred, events_gt, Trange)

which gives as output:

   {'precision': 0.82,
    'recall': 0.84,
    'individual_precision_probabilities': [0.63, 1.0],
    'individual_recall_probabilities': [0.82, 0.87],
    'individual_precision_distances': [0.5, 0.0],
    'individual_recall_distances': [0.5, 0.33]}

Testing and reproducibility

The unit tests can be run by typing:

    python -m unittest discover

The results from the paper are also tested. The specific tests of the results are located at tests/test_data.py and tested against data located in the folder data/.