1. Home
  2. Docs
  3. Fleet-Health
  4. Issue detection
  5. Issue type “Time Outlier”

Issue type “Time Outlier”

What does this issue type detect?

This issue detects a significant difference of time interval between two messages, compared to other messages on that day.

How does this issue type detect issues?

The algorithm first filters out messages where the vehicle is not in motion. After this step, the algorithm determines whether there is enough data to continue analysis. In the case where there is sufficient data, the algorithm uses this data to calculate a threshold. This calculated threshold is dynamic and will be different depending on the data of that day. If the data shows an average time interval of 1 minute between messages, a time interval of 9 minutes will likely not trigger an issue. But if a device usually sends a message every 2 seconds, and suddenly the difference in time between two messages happens to be 9 minutes, this will probably trigger an issue.

If the threshold setting (this threshold can be set in the calculation settings page) is higher that the threshold calculated by the formula, the threshold in the settings is used instead. Any difference in time between two messages higher than the threshold(calculated by the formula or hard threshold set in the calculation settings page), will trigger an issue. The formula used to calculate the dynamic threshold is a modified version of a common formula known in the discipline of statistics as upper fence detection.

What does the threshold setting for this issue type do?

The threshold (in seconds) for this issue type is a setting where issues should not be triggered below the given threshold.

For example an asset might report 5-10 second delay for messages throughout the day, and for just one instance send two messages with a 2 minute delay. The dynamic threshold for this data might be 1 minute. Comparing a 2 minute delay to 5 seconds is indeed a significant difference (24 times as much). But with the threshold in the settings we can override dynamic threshold that are simply too low. We might not want to trigger a Time Outlier issue when the device report a 2 minute delay.

Example of this issue type:

The image below shows a case where in one instance of the day,  the asset was moving and reported two messages that were significantly different from other messages of that day.


Was this article helpful to you? Yes No