Popular Post

Search This Blog

Wednesday, April 18, 2018

Performance #AnomalyDetection Online Course is Launched


What is covered:

  • Machine learning based Anomaly Detection technique
  • Classical (SPC) and MASF (For system performance data) Control Chartirting
  • Where is the Control Chart Used?
  • What are the types of Control Charts?
  • Reading, building, and interpreting Control Charts
  • Typical cases of real world issues captured by anomaly detection system (VMs, Mainframes, Middleware, E2E response and more)
  • How to build free AWS cloud server with R and build there control charts
  • Performance anomaly (Perfomaly) detection system R implementation example (SEDS-lite - open source based tool)
Previews:


"Perfomaly"=Performance Anomaly:



Control Chart as a Machine Learning Tool to Detect Anomalies





See mode details in another post of this blog:

Performance Anomaly ("Perfomaly") Detection. Parts 1-4: Power of Control Charts


Feedback is welcome.



Tuesday, April 17, 2018

TrendieR is a #ChangePointsDetection system based on Exception Value (EV) data analysis

The right term for SETDS methodology is Anomaly and Change (points) detection.

The second part of SETDS - trend detection recently implemented on R and got the "TrendieR" name - is actually a Change Points Detection tool.

See more details how it works in the previous post:

AnomalyDetection vs. NoveltyDetection. SETDS Method Detects and Separates both


But in short the change points are the the following equation roots (solutions)

                  EV(t)=0

Where t is time and EV is Exception Value (magnitude of exception calculated as the difference between actual data and baseline statistical limits UCL and LCL). 

CMG.org paper there was the attempt to describe that calculation more formally by the following formulas:







Wednesday, February 7, 2018

More about #CloudCapacityPlanning from #CMGnews

Visiting #CapitalOneCafe

What is Capacity Management?



What is Capacity Management? [Webinar Recap]: Capacity management is the practice of making sure IT resources meet business demands today and down the road—without over-provisioning. But the role of capacity management has changed as IT environments have evolved.

Tuesday, February 6, 2018

"Machine Learning for Predictive Performance Monitoring" - interesting #CMGJournal article (#CMGnews)

Tim Browning has a lot of good publications about Capacity Management in www.CMG.org  and also in this blog:

"Entropy-Based Anomaly Detection for SAP z/OS Systems"


#CMGamplify - "#DataScience Tools for Infrastructure Operational Intelligence"



"the review of cloud computing article "Optimal Density of Workload Placement"


He has just published his new paper in the CMG Journal:  
            "Machine Learning for Predictive Performance Monitoring",
which is available for CMG members

I have enjoyed reading the paper, below is the abstract:


I like especially his following very true saying: 

"...Machines don’t actually “learn” nor do statistical algorithms represent some mechanistic disembodied intelligence. However, human learning and intelligence is greatly assisted by statistical modeling in much the same way that optics technology assists vision..."

I appreciate he referenced two my CMG papers under his "Useful Related Materials" section:

- Trubin, Igor, “Exception Based Modeling and Forecasting”, CMG2008 Proceedings
- Trubin, Igor, “Capturing Workload Pathology by Statistical Exception Detection System”,
CMG2005 Proceedings.





Thursday, February 1, 2018

#AnomalyDetection vs. #NoveltyDetection. SETDS Method Detects and Separates both

Reading "Anomaly detection with Apache MXNet":

"An important distinction has to be made between anomaly detection and “novelty detection.” The latter turns up new, previously unobserved, events that still are acceptable and expected. For example, at some point in time, your credit card statements might start showing baby products, which you’ve never before purchased. Those are new observations not found in the training data, but given the normal changes in consumers’ lives, may be acceptable purchases that should not be marked as anomalies."

I figured out that my SETDS method has this Novelty Detection included as my

EV based trends detection  method (e.g. implemented in R as "TrendieR") finds recent change points in the time-serious data and then by building trend-forecast checks if the change is permanent or not. So if it is permanent the possible "novelty" is detected.  



So the 1st part of SETDS  (e.g. implemented as "SonR" on R) captures just anomalies and/or outliers, then Trend detection separates cases that indicate the possible "novelty". (something changed and stays changed and growing). Still false positive could be there though.... 

BTW there is a 3rd level of SETDS which is actually the way to correlate performance data with demand (drivers) data  to build meaningful forecasts (e.g. implemented as "Model Factory")