Anomaly Detection in IoT
The last couple of years, the amount of devices connected to the Internet has increased dramatically, usually referred to as the “Internet of Things”. Connecting devices opens up opportunities for real-time access and remote controlling, but the access to data from these devices also enables more sophisticated research and machine learning based functionality. Most likely, this data is structured as a time series, and many times both fine-grained and multivariate.
In the IoT domain, there are a multitude of applications for machine learning and AI, ranging from forecasting to segmentation. A common problem is the need to predict uncommon events and it is often so that the cause is unclear. This may include for example equipment breakdowns, fraudulent behavior, health monitoring, etc. All of these are examples of situations where it can be fruitful to look for strange/novel behavior in the data leading up to the event in question.
The session starts with a detailed walkthrough of the problem domain, focusing on different approaches and techniques for anomaly detection, primarily for time series data. Afterwards, the attendees will have a choice of either getting hands-on with one or more of these techniques, or to engage in a discussion on how anomaly becomes relevant in a practical scenario.
About the Presenters:
Rasmus Thornberg has an education in Engineering Physics from Lund University and has worked for eon, Sony, and Assa Abloy, Rasmus is currently Senior Data Scientist at Sigma IT Consulting.
Mattias Jönsson has an education in Informatics from Lund University and has worked for IBM and Microsoft. Mattias is currently Senior Data Scientist at Sigma IT Consulting.
If you interested in presenting for HandOn Data Science please contact Philippe Wee at firstname.lastname@example.org
17.30-17.45 Meet & Greet
17.45-18.30 Introduction to Anomaly Detection
19.00-19.45 Hands-on Labs or Group Discussion
19.45-20.00 Wrap-up and conclusions