Failure prediction in real time on time series data can be realized with the use of Open Source tools. We will deliver an overall view of how to start with the generation of new raw sensor data (typically captured by an Edge device), and end up with a real time graph that shows alerts warning that a failure is imminent. There are a number of processes that must be put into place before the stated goal can be fully realized.

In particular, we will first need a data collector whose job is to receive the raw sensor data and then put that data into a data storage unit. All of the new raw sensor data are associated with the timestamp of when the sensor data point was generated, thereby forming what is called a time series. The data collector then, by socket communication sends the new data to a web application that puts the new data into a form that enables a trained Machine Learning model to make a binary classification (Normal or Not Normal) prediction. A real time time series graph is then updated with the prediction, and the data is pushed to a browser where the real time graph is rendered. All of the processes mentioned above can be implemented with the use of Open Source tooling.

Local ODSC chapter in NYC, USA

Instructor's Bio

Audrey Reznik

Sr. Principal Software Engineer at Red Hat

Audrey is part of Red Hat OpenShift Data Science team focusing on helping customers with managed services, AI/ML workloads and next-generation platforms. She holds degrees in Computer Information Systems and Geology, and has work experience in each field. Audrey is passionate about Data Science and in particular the current opportunities with AIML at the Edge and Open Source technologies.


  • 1

    ON-DEMAND WEBINAR: Using Open Source for Failure Prediction

    • Ai+ Training

    • Webinar Recording

    • Welcome to ODSC East 2023 in Boston or virtually!