In this video, Romeo Kienzler walks through a Node-RED flow that calculates a moving z-score to detect anomalies on simulated data and that visualizes that data in a dashboard.

In this video:

After explaining how the moving z-score is calculated with means and standard deviations for all values in a sliding window, Romeo demonstrates how to implement this in a Node-RED app.

Romeo shows how to install the Node-RED dashboard package, which includes many new nodes for building node apps. Then, he shows how to import an existing flow from a JSON file in the defaults directory of his Cognitive IoT GitHub repo.

Romeo then talks you through the various nodes in this Node-RED flow, to show you how you can detect anomolies using the moving z-score. This flow can be run on any device on the IoT edge, allowing the analytics to occur on the IoT edge. The data is captured and sent to the Watson IoT Platform using MQTT for further downstream analytics. Finally, he walks through the function node that contains the code for calculating the moving z-score. He shows the two chart nodes from the dashboard package that shows the data visualization on the moving z-score and the thresholds that are set.

Discovering Data Science with Romeo Kienzler

Follow Romeo as he tackles the most difficult challenges in data science.

1 comment on"Detect anomalies using moving z-score on the IoT edge using Node-RED"

  1. NAPOLEAO PERES FIGUEIRA December 20, 2018


Join The Discussion

Your email address will not be published. Required fields are marked *