===AIChannel===

The **AIChannel** function of the [[mplreport]] class writes an array of data to a file for AI analysis.

This is used by the reporting engine to help evaluate what a 'normal' report looks like, so that unusual or unexpected events in reports can be highlighted to users.

==Parameters==

^Parameter^Type^Default^Meaning^
|section|str||The name of the report section. If you have multiple axes in your report, it might be helpful to split the AI analysis up across each axis.|
|arr|list||The data to be analysed. This must be an array of floating point numbers. If possible, they should be //normalised// between 0 and 1 or -1 and 1.|

==Return==

None

==Example==

<code python>
totals = {}
totals.append(df['Section A.Power'].sum())
totals.append(df['Section B.Power'].sum())

report.AIChannel("Power Totals",totals)
</code>

==Usage==

The reporting system prompts the user to confirm if the report they are looking at is 'normal' or not. 

If its reported as looking normal, the numbers that were passed to this function are recorded into training data used to build an anomaly detector. 

You pass a name along with the array because some reports might have multiple different types of KPI being calculated in a single report. By splitting them into distinctly-named arrays, the system can be much more specific about the nature of the problem.

For example, a report might calculate //total production//, //total energy use// and //total gas use// per section of your site. By having these in three distinct named arrays, the anomaly detection system can tell you that there's an anomaly specifically in the //total energy// calculations.