Cross-Industry Anomaly Detection Solutions with R and Shiny
summary
Demonstrates production anomaly detection using the Twitter-derived R package (forked to return tibbles) to catch hardware failures and news spikes, combining piecewise median decomposition, outlier filtering, and simple business-rule limits to reduce false positives. Prototype built as a real-time Shiny app that connects to live databases, uses one year of data for seasonality, lets users tune significance and quartile-based thresholds, and is deployed daily to successfully surface failures and newsworthy spikes and to catch tagging errors.
AI-written summary · about
related talks
18mCross-Industry Anomaly Detection Solutions with R and Shiny
42mDetecting novel anomalies in Twitter
01h 16mAnomaly Detection
30mBetter data quality with ML & visualization
39mMastering real-time anomaly detection with open source tools
29mREAL-TIME ANOMALY DETECTION MADE EASY
14mChaos Genius: Open Sourcing Outlier Detection and Root Cause Analysis
45mAnomaly Detection in Time Series