I am pleased to introduce an integration between IBM Watson IOT platform and IBM Predictive analytic as a service to provide an machine learning algorithm that detects anomalous behavior of things. In a typical Industrial deployment, Anomaly detection is the most sought after algorithm that can forecast trends and detects data that violates the pattern. These insights are leveraged by the IBM Watson IOT Platform to create rules for Alerts and visualizations for dashboards for business users to act upon.

The Recipe

Titled “Engage Machine Learning for detecting anomalous behaviors of things” provides the following capabilities:

  • The recipe provides an out of the model predictive model that detects spikes and dips from the IOT real time data.
  • The model identifies the normal pattern of behavior from the real time events and associates an appropriate predictive model.
  • The Model is limited to Univariate time series analysis.
  • Predictive Analytics as a service provides a bluemix based cloud service to host SPSS models.
  • Though the instructions are tuned to engage the distributed SPSS model, one could easily replace the model with their custom streams.

I encourage you to incorporate Anomaly detection in your IOT missions and would love to hear your feedback.

2 comments on"Engage Machine Learning models to detect anomalous behavior of things"

  1. Frank Williams April 12, 2016

    Does Watson identifies visual anomalies?

Join The Discussion

Your email address will not be published.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