Developing "spidey sense"
summary
Demo runs on a Raspberry Pi with a Sense HAT, Node (Jupyter/tslab) and the stats-analysis npm, collecting minutely time series and sending batches to Azure Anomaly Detector via the Node SDK; the API requires a minimum of 12 and supports up to 8,640 entries and returns a last-point anomaly flag. The demo forces a spike to prove detection, then lights an LED X on the Sense HAT when the point is anomalous.
AI-written summary · about
chapters (16)
- 0:00Developing spidey sense
- 3:24Anomaly detection overview
- 6:15Rule-based anomaly detection
- 8:03Statistical anomaly detection
- 9:31Machine learning approaches for detection
- 11:38JS anomaly detection demo
- 12:00Time series data
- 13:23Time series anomalies
- 15:56Univariate vs multivariate
- 16:39Azure anomaly detector
- 20:35Setup anomaly detector API
- 21:46Raspberry Pi sensor data
- 24:15Sending data to detector
- 27:45Anomaly alerts
- 29:52Applications and integrations
- 30:57Summary
related talks
39mDeveloping Spidey Senses
52mDevil in the Haystack
28mReal Time Sensor Anomaly Detection with Sci Kit Learn and the Azure Stack
35mObservability’s Sixth Sense: Detecting Anomalies in Metrics
37mPracticing Data Science in the cloud
19mReal-time Experiments with an AI Co-Scientist
29mAnomaly Detection
01h 00mIntelligent Interfaces with Cognitive APIs, BOTs and IOT