Editor’s Note: We have known the Anodot team for a few years, and have been following the development of their product closely. If you have ever looked at logs or long lists of recordings, you know how hard it is to spot slight variations, or even major anomalies efficiently. In the world of massive quantities of data generated by sensors everywhere, the ability to intelligently identify trends, seasonality and the real outliers is critical. Anodot does it simply and elegantly. In this guest blog post, the Anodot team introduces their business intelligence solution and how it works to provide incident detection with ARTIK Cloud.
When talking about IoT, it can be easy to get carried away with visions of refrigerators that automatically order milk when it runs out, autonomous cars that save fuel and avoid accidents, and factories that manufacture and ship products without a single human interaction.
However, no matter how interconnected machines become, human beings still need a way to monitor them, to check that everything is working as planned. If refrigerators suddenly overload themselves with milk, for example, or autonomous cars suddenly start driving off the road, there needs to be a system in place to identify these unusual situations.
The vast majority of the potential value of IoT requires the ability to integrate and analyze data from multiple IoT systems, according to a McKinsey report. With its open APIs, Samsung ARTIK Cloud brings this data together into one place. By connecting directly to ARTIK Cloud, Anodot enables a layer of analytics and real-time detection on top of the collected data.
Anodot: Automated Machine Learning of Your Metrics
Anodot makes sense of the millions of data points streaming into ARTIK Cloud from the various IoT sensors in your home, factory or other IoT implementation. Anodot is disrupting the traditional business intelligence industry with its real time business incident detection solution.
Anodot uses proprietary machine learning algorithms to learn the normal behavior of the data in ARTIK Cloud, and then sends alerts about any metrics that behave in an abnormal way. Anodot scores each anomaly based on how far “off” it is from the normal, and correlates multiple related anomalies to avoid alert storms and aid in determining the root cause of any issue that arises.
While Anodot could be considered a business intelligence (BI) solution, it differs in several key areas:
- BI looks backward at historical data, Anodot analyzes streaming data in real time and predicts the future behavior of each metric.
- BI needs the user to decide what questions to ask and what data to look at; Anodot automatically identifies what is happening that you need to see.
- BI typically focuses on a few key metrics; Anodot can ingest all of your metrics, and then focus on just the important ones.
Predictive Maintenance Prevents Breakdowns
In the screenshot below, a factory is using Anodot to monitor the data generated by its sensors. All machine parameters are tracked and learned in real time, including material consumption, RPM, temperature, total errors and machine output. In the screenshot below, Anodot has identified a drop in throughput and an increase in material consumption, possibly indicating that the machine has a problem.
Similarly, in a smart home, Anodot can track all the data generated by home sensors. If the washing machine is suddenly taking longer to wash, or the air conditioning is coming on but temperature stays constant, these types of anomalies will trigger an alert which the homeowner can investigate.
Outlier Detection Identifies When Maintenance is Necessary
Anodot can also be used to compare the performance of similar things. For example, in a home with ten of the same light fixtures, if one is heating up more than the others, it could indicate a problem in a specific fixture.
In an industrial scenario, when a company has multiple machines operating in a similar way, Anodot can spot outliers not just within the data of one machine, but by correlating data from multiple machines (where the same input, temperature, etc., might be expected).
Make Sense of Your ARTIK Cloud Data
Diverse devices and “things” send and receive messages to ARTIK Cloud continuously. Anodot helps you make sense of what is happening in real time. You can receive automated, preemptive alerts, without having to set any thresholds or even understand how the data is supposed to behave.
Often machine data has inherent patterns, for example lights that are turned on in the evening, turned off at night, and then on again in the early morning. On the weekends the behavior may be different. These and other patterns make it difficult for a human user to identify static thresholds to set for manual alerts, and in fact make most static thresholds irrelevant. Anodot, however, automatically learns how the data behaves, including all of its patterns and seemingly random behavior. It can also correlate between external events (for example, holidays or weather changes) and the data metrics.
Try Anodot for Free