===AI Report Analysis===

ARDI can analyse your report output and - with some training from your users - use machine learning to tell the difference between a 'normal' and an 'abnormal' report.

This helps combat //report fatigue// - people skimming over reports without really reading them - by highlighting those reports that appear to have unusual information.

==Adding AI to your Report==

We don't suggest pushing large amounts of data for AI analysis - sending noisy raw data will usually not give you a lot of value.

However, summary information such as minimum, maximum and average measurements can be //extremely useful// in AI analysis. 

To use AI analysis, write a JSON file containing a dictionary of data point arrays. This file should have the //same name as your report//, but with the **.json** file extension.

For example, if we wanted to record the minimum, maximum and average value(s) of our temperature data, our file should look like the one below...

<code>
{
   "Temperatures": [80,200,185]
}
</code>

For multi-channel reports, you might choose to have several different items in the dictionary - this helps the AI explain //which part// of the report is unusual when you have a lot of data shown on the one page.

<code>
{
   "Temperatures": [80,200,185],
   "Pressures": [2000,2500,3000]
}
</code>

If you're using our [[mplreport:welcome|reporting library]], you can do this automatically using the [[mplreport:aichannel|AIChannel]] function.