Demo: Detect malfunctioning IoT sensors
Imagine that you have data from multiple Internet-of-Things devices like sensors, connected cars, or other smart devices. At times, you might need to correlate the data from multiple devices to make decisions or detect anomalies. To do this, you’d need to send the data from all the devices to a central application for more processing.
If your central application allows you to analyze the data as it is received, instead of requiring you to store it first, you can extract insights in real time. You could also check the data for correctness before usage. For instance, you could flag any devices that are reporting data that is unusual when compared to the general average. Such a solution prevents you from storing erroneous data. It would also save the cost of physically testing each device.
You can create and deploy such an application right in your browser with the products in the Watson Data Platform. Using a Python notebook in the Data Science Experience and the IBM Streams Python API, you can ingest and analyze your IoT data. You could then send the results for visualization or storage to other services like PixieDust, Plotly, Cloudant, or MessageHub.
This short video features a Streams application that analyzes data from IoT devices and does some basic outlier detection to determine if any of them is malfunctioning. The results are updated on a map with markers indicating which devices are reporting suspicious data.