Skip to main content

No project description provided

Project description

Precision and Recall for Time Series

License Python package PyPI version

Unofficial python implementation of Precision and Recall for Time Series.

Classical anomaly detection is principally concerned with point-based anomalies, those anomalies that occur at a single point in time. Yet, many real-world anomalies are range-based, meaning they occur over a period of time. Motivated by this observation, we present a new mathematical model to evaluate the accuracy of time series classification algorithms. Our model expands the well-known Precision and Recall metrics to measure ranges, while simultaneously enabling customization support for domain-specific preferences.

This is the open source software released by Computational Mathematics Laboratory. It is available for download on PyPI.



PRTS is on PyPI, so you can use pip to install it.

$ pip install prts

from github

You can also use the following command to install.

$ git clone
$ cd PRTS
$ make install  # (or make develop)


from prts import ts_precision, ts_recall

# calculate time series precision score
precision_flat = ts_precision(real, pred, alpha=0.0, cardinality="reciprocal", bias="flat")
precision_front = ts_precision(real, pred, alpha=0.0, cardinality="reciprocal", bias="front")
precision_middle = ts_precision(real, pred, alpha=0.0, cardinality="reciprocal", bias="middle")
precision_back = ts_precision(real, pred, alpha=0.0, cardinality="reciprocal", bias="back")
print("precision_flat=", precision_flat)
print("precision_front=", precision_front)
print("precision_middle=", precision_middle)
print("precision_back=", precision_back)

# calculate time series recall score
recall_flat = ts_recall(real, pred, alpha=0.0, cardinality="reciprocal", bias="flat")
recall_front = ts_recall(real, pred, alpha=0.0, cardinality="reciprocal", bias="front")
recall_middle = ts_recall(real, pred, alpha=0.0, cardinality="reciprocal", bias="middle")
recall_back = ts_recall(real, pred, alpha=0.0, cardinality="reciprocal", bias="back")
print("recall_flat=", recall_flat)
print("recall_front=", recall_front)
print("recall_middle=", recall_middle)
print("recall_back=", recall_back)


Parameter Description Type
alpha Relative importance of existence reward (0 ≤ alpha ≤ 1). float
cardinality Cardinality type. This should be "one", "reciprocal" or "udf_gamma" string
bias Positional bias. This should be "flat", "front", "middle", or "back" string


We provide a simple example code. By the following command you can run the example code for the toy dataset and visualize the metrics.

$ python3 examples/

example output


You can run all the test codes as follows:

$ make test


  • Tatbul, Nesime, Tae Jun Lee, Stan Zdonik, Mejbah Alam, and Justin Gottschlich. 2018. “Precision and Recall for Time Series.” In Advances in Neural Information Processing Systems, edited by S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, 31:1920–30. Curran Associates, Inc.


This repository is Apache-style licensed, as found in the LICENSE file.


  doi = {10.5281/ZENODO.4428056},
  url = {},
  author = {Ryohei Izawa, Ryosuke Sato, Masanari Kimura},
  title = {PRTS: Python Library for Time Series Metrics},
  publisher = {Zenodo},
  year = {2021},
  copyright = {Open Access}

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

prts- (12.7 kB view hashes)

Uploaded source

Built Distribution

prts- (13.8 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page