Anomaly Detection and Correlation library

Overview

luminol

Python Versions Build Status

Overview

Luminol is a light weight python library for time series data analysis. The two major functionalities it supports are anomaly detection and correlation. It can be used to investigate possible causes of anomaly. You collect time series data and Luminol can:

  • Given a time series, detect if the data contains any anomaly and gives you back a time window where the anomaly happened in, a time stamp where the anomaly reaches its severity, and a score indicating how severe is the anomaly compare to others in the time series.
  • Given two time series, help find their correlation coefficient. Since the correlation mechanism allows a shift room, you are able to correlate two peaks that are slightly apart in time.

Luminol is configurable in a sense that you can choose which specific algorithm you want to use for anomaly detection or correlation. In addition, the library does not rely on any predefined threshold on the values of a time series. Instead, it assigns each data point an anomaly score and identifies anomalies using the scores.

By using the library, we can establish a logic flow for root cause analysis. For example, suppose there is a spike in network latency:

  • Anomaly detection discovers the spike in network latency time series
  • Get the anomaly period of the spike, and correlate with other system metrics(GC, IO, CPU, etc.) in the same time range
  • Get a ranked list of correlated metrics, and the root cause candidates are likely to be on the top.

Investigating the possible ways to automate root cause analysis is one of the main reasons we developed this library and it will be a fundamental part of the future work.


Installation

make sure you have python, pip, numpy, and install directly through pip:

pip install luminol

the most up-to-date version of the library is 0.4.


Quick Start

This is a quick start guide for using luminol for time series analysis.

  1. import the library
import luminol
  1. conduct anomaly detection on a single time series ts.
detector = luminol.anomaly_detector.AnomalyDetector(ts)
anomalies = detector.get_anomalies()
  1. if there is anomaly, correlate the first anomaly period with a secondary time series ts2.
if anomalies:
    time_period = anomalies[0].get_time_window()
    correlator = luminol.correlator.Correlator(ts, ts2, time_period)
  1. print the correlation coefficient
print(correlator.get_correlation_result().coefficient)

These are really simple use of luminol. For information about the parameter types, return types and optional parameters, please refer to the API.


Modules

Modules in Luminol refers to customized classes developed for better data representation, which are Anomaly, CorrelationResult and TimeSeries.

Anomaly

class luminol.modules.anomaly.Anomaly
It contains these attributes:

self.start_timestamp: # epoch seconds represents the start of the anomaly period.
self.end_timestamp: # epoch seconds represents the end of the anomaly period.
self.anomaly_score: # a score indicating how severe is this anomaly.
self.exact_timestamp: # epoch seconds indicates when the anomaly reaches its severity.

It has these public methods:

  • get_time_window(): returns a tuple (start_timestamp, end_timestamp).

CorrelationResult

class luminol.modules.correlation_result.CorrelationResult
It contains these attributes:

self.coefficient: # correlation coefficient.
self.shift: # the amount of shift needed to get the above coefficient.
self.shifted_coefficient: # a correlation coefficient with shift taken into account.

TimeSeries

class luminol.modules.time_series.TimeSeries

__init__(self, series)
  • series(dict): timestamp -> value

It has a various handy methods for manipulating time series, including generator iterkeys, itervalues, and iteritems. It also supports binary operations such as add and subtract. Please refer to the code and inline comments for more information.


API

The library contains two classes: AnomalyDetector and Correlator, and there are two sets of APIs, one corresponding to each class. There are also customized modules for better data representation. The Modules section in this documentation may provide useful information as you walk through the APIs.

AnomalyDetector

class luminol.anomaly_detector.AnomalyDetecor

__init__(self, time_series, baseline_time_series=None, score_only=False, score_threshold=None,
         score_percentile_threshold=None, algorithm_name=None, algorithm_params=None,
         refine_algorithm_name=None, refine_algorithm_params=None)
  • time_series: The metric you want to conduct anomaly detection on. It can have the following three types:
1. string: # path to a csv file
2. dict: # timestamp -> value
3. lumnol.modules.time_series.TimeSeries
  • baseline_time_series: an optional baseline time series of one the types mentioned above.
  • score only(bool): if asserted, anomaly scores for the time series will be available, while anomaly periods will not be identified.
  • score_threshold: if passed, anomaly scores above this value will be identified as anomaly. It can override score_percentile_threshold.
  • score_precentile_threshold: if passed, anomaly scores above this percentile will be identified as anomaly. It can not override score_threshold.
  • algorithm_name(string): if passed, the specific algorithm will be used to compute anomaly scores.
  • algorithm_params(dict): additional parameters for algorithm specified by algorithm_name.
  • refine_algorithm_name(string): if passed, the specific algorithm will be used to compute the time stamp of severity within each anomaly period.
  • refine_algorithm_params(dict): additional parameters for algorithm specified by refine_algorithm_params.

Available algorithms and their additional parameters are:

1.  'bitmap_detector': # behaves well for huge data sets, and it is the default detector.
    {
      'precision'(4): # how many sections to categorize values,
      'lag_window_size'(2% of the series length): # lagging window size,
      'future_window_size'(2% of the series length): # future window size,
      'chunk_size'(2): # chunk size.
    }
2.  'default_detector': # used when other algorithms fails, not meant to be explicitly used.
3.  'derivative_detector': # meant to be used when abrupt changes of value are of main interest.
    {
      'smoothing factor'(0.2): # smoothing factor used to compute exponential moving averages
                                # of derivatives.
    }
4.  'exp_avg_detector': # meant to be used when values are in a roughly stationary range.
                        # and it is the default refine algorithm.
    {
      'smoothing factor'(0.2): # smoothing factor used to compute exponential moving averages.
      'lag_window_size'(20% of the series length): # lagging window size.
      'use_lag_window'(False): # if asserted, a lagging window of size lag_window_size will be used.
    }

It may seem vague for the meanings of some parameters above. Here are some useful insights:

The AnomalyDetector class has the following public methods:

  • get_all_scores(): returns an anomaly score time series of type TimeSeries.
  • get_anomalies(): return a list of Anomaly objects.

Correlator

class luminol.correlator.Correlator

__init__(self, time_series_a, time_series_b, time_period=None, use_anomaly_score=False,
         algorithm_name=None, algorithm_params=None)
  • time_series_a: a time series, for its type, please refer to time_series for AnomalyDetector above.
  • time_series_b: a time series, for its type, please refer to time_series for AnomalyDetector above.
  • time_period(tuple): a time period where to correlate the two time series.
  • use_anomaly_score(bool): if asserted, the anomaly scores of the time series will be used to compute correlation coefficient instead of the original data in the time series.
  • algorithm_name: if passed, the specific algorithm will be used to calculate correlation coefficient.
  • algorithm_params: any additional parameters for the algorithm specified by algorithm_name.

Available algorithms and their additional parameters are:

1.  'cross_correlator': # when correlate two time series, it tries to shift the series around so that it
                       # can catch spikes that are slightly apart in time.
    {
      'max_shift_seconds'(60): # maximal allowed shift room in seconds,
      'shift_impact'(0.05): # weight of shift in the shifted coefficient.
    }

The Correlator class has the following public methods:

  • get_correlation_result(): return a CorrelationResult object.
  • is_correlated(threshold=0.7): if coefficient above the passed in threshold, return a CorrelationResult object. Otherwise, return false.

Example

  1. Calculate anomaly scores.
from luminol.anomaly_detector import AnomalyDetector

ts = {0: 0, 1: 0.5, 2: 1, 3: 1, 4: 1, 5: 0, 6: 0, 7: 0, 8: 0}

my_detector = AnomalyDetector(ts)
score = my_detector.get_all_scores()
for timestamp, value in score.iteritems():
    print(timestamp, value)

""" Output:
0 0.0
1 0.873128250131
2 1.57163085024
3 2.13633686334
4 1.70906949067
5 2.90541813415
6 1.17154110935
7 0.937232887479
8 0.749786309983
"""
  1. Correlate ts1 with ts2 on every anomaly.
from luminol.anomaly_detector import AnomalyDetector
from luminol.correlator import Correlator

ts1 = {0: 0, 1: 0.5, 2: 1, 3: 1, 4: 1, 5: 0, 6: 0, 7: 0, 8: 0}
ts2 = {0: 0, 1: 0.5, 2: 1, 3: 0.5, 4: 1, 5: 0, 6: 1, 7: 1, 8: 1}

my_detector = AnomalyDetector(ts1, score_threshold=1.5)
score = my_detector.get_all_scores()
anomalies = my_detector.get_anomalies()
for a in anomalies:
    time_period = a.get_time_window()
    my_correlator = Correlator(ts1, ts2, time_period)
    if my_correlator.is_correlated(threshold=0.8):
        print("ts2 correlate with ts1 at time period (%d, %d)" % time_period)

""" Output:
ts2 correlates with ts1 at time period (2, 5)
"""

Contributing

Clone source and install package and dev requirements:

pip install -r requirements.txt
pip install pytest pytest-cov pylama

Tests and linting run with:

python -m pytest --cov=src/luminol/ src/luminol/tests/
python -m pylama -i E501 src/luminol/
Owner
LinkedIn
LinkedIn
A library to generate synthetic time series data by easy-to-use factors and generator

timeseries-generator This repository consists of a python packages that generates synthetic time series dataset in a generic way (under /timeseries_ge

Nike Inc. 87 Dec 20, 2022
database for artificial intelligence/machine learning data

AIDB v0.0.1 database for artificial intelligence/machine learning data Overview aidb is a database designed for large dataset for machine learning pro

Aarush Gupta 1 Oct 24, 2021
A repository to index and organize the latest machine learning courses found on YouTube.

📺 ML YouTube Courses At DAIR.AI we ❤️ open education. We are excited to share some of the best and most recent machine learning courses available on

DAIR.AI 9.6k Jan 01, 2023
To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction

To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction. The challenge aims to adress the problems of medical imbalanced data classification.

Marwan Mashra 1 Jan 31, 2022
Case studies with Bayesian methods

Case studies with Bayesian methods

Baze Petrushev 8 Nov 26, 2022
Unofficial pytorch implementation of the paper "Context Reasoning Attention Network for Image Super-Resolution (ICCV 2021)"

CRAN Unofficial pytorch implementation of the paper "Context Reasoning Attention Network for Image Super-Resolution (ICCV 2021)" This code doesn't exa

4 Nov 11, 2021
A python library for easy manipulation and forecasting of time series.

Time Series Made Easy in Python darts is a python library for easy manipulation and forecasting of time series. It contains a variety of models, from

Unit8 5.2k Jan 04, 2023
Predict the income for each percentile of the population (Python) - FRENCH

05.income-prediction Predict the income for each percentile of the population (Python) - FRENCH Effectuez une prédiction de revenus Prérequis Pour ce

1 Feb 13, 2022
Accelerating model creation and evaluation.

EmeraldML A machine learning library for streamlining the process of (1) cleaning and splitting data, (2) training, optimizing, and testing various mo

Yusuf 0 Dec 06, 2021
A project based example of Data pipelines, ML workflow management, API endpoints and Monitoring.

MLOps template with examples for Data pipelines, ML workflow management, API development and Monitoring.

Utsav 33 Dec 03, 2022
K-means clustering is a method used for clustering analysis, especially in data mining and statistics.

K Means Algorithm What is K Means This algorithm is an iterative algorithm that partitions the dataset according to their features into K number of pr

1 Nov 01, 2021
🌲 Implementation of the Robust Random Cut Forest algorithm for anomaly detection on streams

🌲 Implementation of the Robust Random Cut Forest algorithm for anomaly detection on streams

Real-time water systems lab 416 Jan 06, 2023
Primitives for machine learning and data science.

An Open Source Project from the Data to AI Lab, at MIT MLPrimitives Pipelines and primitives for machine learning and data science. Documentation: htt

MLBazaar 65 Dec 29, 2022
learn python in 100 days, a simple step could be follow from beginner to master of every aspect of python programming and project also include side project which you can use as demo project for your personal portfolio

learn python in 100 days, a simple step could be follow from beginner to master of every aspect of python programming and project also include side project which you can use as demo project for your

BDFD 6 Nov 05, 2022
Machine-Learning with python (jupyter)

Machine-Learning with python (jupyter) 머신러닝 야학 작심 10일과 쥬피터 노트북 기반 데이터 사이언스 시작 들어가기전 https://nbviewer.org/ 페이지를 통해서 쥬피터 노트북 내용을 볼 수 있다. 위 페이지에서 현재 레포 기

HyeonWoo Jeong 1 Jan 23, 2022
GRaNDPapA: Generator of Rad Names from Decent Paper Acronyms

Generator of Rad Names from Decent Paper Acronyms

264 Nov 08, 2022
WAGMA-SGD is a decentralized asynchronous SGD for distributed deep learning training based on model averaging.

WAGMA-SGD is a decentralized asynchronous SGD based on wait-avoiding group model averaging. The synchronization is relaxed by making the collectives externally-triggerable, namely, a collective can b

Shigang Li 6 Jun 18, 2022
CorrProxies - Optimizing Machine Learning Inference Queries with Correlative Proxy Models

CorrProxies - Optimizing Machine Learning Inference Queries with Correlative Proxy Models

ZhihuiYangCS 8 Jun 07, 2022
Predicting job salaries from ads - a Kaggle competition

Predicting job salaries from ads - a Kaggle competition

Zygmunt Zając 57 Oct 23, 2020
ETNA is an easy-to-use time series forecasting framework.

ETNA is an easy-to-use time series forecasting framework. It includes built in toolkits for time series preprocessing, feature generation, a variety of predictive models with unified interface - from

Tinkoff.AI 674 Jan 07, 2023