Robust anomaly detection for real user monitoring data - Velocity 2016, Santa Clara, CA

[@arnamoy10]

Loved the use of shapes for algo and anomalies to explain FP and FNs, will steal the idea for my presentation ;)

[@PavanKumar-hp1el]

Can anyone suggest some research paper on luminol, so that proper algorithm what it tells I can learn

[@ashwinikadam9002]

Hi Ritesh Maheshwari, i wanted to learn about luminol library, can you please suggest some tutorials, or wanted to learn you if have some time to teach.

[@Josh-ct5uk]

I am researching about anomallies detection, I also watched your clip but I do not know that how we generate baseline from time series dataset to build our detect model?

[@amitgupta-dy4vg]

Hi Toan, I am working on the same anomaly detection problem. I thought it would be nice to connect, tried searching you on linkedin but couldn't find you. Can you please ping me at linkedin.com/in/scanamit

[@jiaqisun6016]

Hi I am creating the same system.I sent an invitation to u on linkedin plz check it.

[@hrushikeshkulkarni8118]

Baseline is basically, you can have avg metric value for last x days.. This will give you trend

[@naveena1580]

bro, any reason why not a classification algorithm to identify root cause instead of an if-else?

[@astroganov]

Because someone should prepare a training dataset to fit your classification algorithm. And that is a huge afford. In this case it's reasonable to use heuristic approach. The second reason is there could be more than one root causes at the same time, and classification model will pick only one for you - the most probable. If-else approach can give you the set of root causes.