Popular Post


Tuesday, February 4, 2020

#SEDS-Lite goes online. New web app to catch perfomalies will be announced at #CMGIMPACT2020 "Meeting of the Minds: AI, ML, and DL"

The SEDS/SETDS method was developed long ago and was implemented as a products/reportings by different folks for different companies. All that is reported and discussed in this blog.

Several years ago I offered to community the way to develop an open-source based tool to use the method and published the following CMG paper about that:

SEDS-Lite: Using Open Source Tools (R, BIRT, MySQL) to Report and Analyze Performance Data 

Finally the group of my friends-developers has started implementing that as a web app. 
At the upcoming CMGIMPACT2020 conference they plan to announce the version v1.0 of the tool, which covers the following:

1. VISUALIZATION. Weekly (Monthly in v2+) data profiling to visualize patterns, anomalies and short term seasonality via IT-Control Charts. (v1.0)
2. ANALYSIS. Anomalies and Change Points Detection in date-time stamped data. (v2+)
INPUTCSV file with timestamp data (time series observations of a dynamic object). The example of the input file can be downloaded form HERE.

Data granularity hourly (v1.0); minutely, daily (v2+)

- IT-Control Chart (see example below)  (v1.0)
- Data cube with  summarized data (168 rows/weekhours - v1.0)
- List of anomalies and change points (v2+)
Input data should consist of at least 3 weeks of history as the method requires comparing the last 7 days of data (actual) with at least 2 weeks long learning/reference data set (baseline). The size of the history is limited by about 5 years. Unlimited size of input data will be implemented in v.2.

Additional resources:
How to read IT-Control Chart (on-line article)
- Online class "Performance Anomaly Detection"

Project contributors:
Anfisa Trubina
- Philip Trubin

If you are interested, please attend the following session (in person or streaming):

Time and Day: 5:15PM Monday, February 10

No comments:

Post a Comment