[Metrics] Common Metrics - Burstiness

X-lab2017 / open-digger

Open source analysis tools

https://open-digger.cn

Apache License 2.0

292 stars 86 forks source link

[Metrics] Common Metrics - Burstiness #572

Open xiaoya-yaya opened 3 years ago

xiaoya-yaya commented 3 years ago

Description:

Short timeframes of intense activity, followed by a corresponding return to a typical pattern of activity, observed in a project.
Burstiness is a way of understanding the cycle of activity in existing metrics, like issues, merge requests, mailing lists, commits, or comments.

Filters

Stars
Forks
Issues or bug reports
Labels
Downloads
Release Tags
Change Requests
Documentation additions or revisions
New Repositories
Feature Requests

Instances of Implementation

Resource

CHAOSS: Common Metrics -> When
https://chaoss.community/metric-burstiness/

open-digger-bot[bot] commented 3 years ago

This issue has not been replied for 24 hours, please pay attention to this issue: @sunshinemingo @wengzhenjie

frank-zsy commented 1 year ago

I can not see what to implement in this metric, the charts just show the event count and the burstiness is just a sense of the smoothness of the charts, is it right? @xiaoya-yaya

xiaoya-yaya commented 1 year ago

I think this metric provides a perspective for maintainers to observe: to find the peaks of the number of commits, comments, downloads, forks, etc, and then to find connections of facts that are related to the burstiness (release, meetups, etc).

The observation could be just based on the observer's sense, at least CHAOSS didn't provide specific methods. However, there are some peak detection methods integrated into python or javascript packages. I know scipy have signal.find_peaks() function to identify peaks in curves.

frank-zsy commented 1 year ago

Thanks for the input, I found a blog about scipy.signal.find_peaks and it gives a corresponding JavaScript implementation.

According to the blog, I think peak detection can be a serious research problem and there are several parameters in the function, so if we want to implement this metric, we may need to look into the peak detection algorithm more carefully.

But it is truly a good metric to build a monitor-and-alert system for other metrics.

birdflyi commented 1 year ago

I am quite curious about the predictability of those signal features:

the max values or relative ratio sequence of peaks,
the interval sequence of peaks,
parameter ratio of fractal if fractal exists.

Is there any set of metrics exists like this?