Extract information from the rules generated by the temporal training

We have a deployment on a large telecom company in Brazil. The end users are used to work with custom tailored correlations, and are worried about enabling AI generated rules without prior human review. However, since the network is very large, one temporal training results in over 100k insights, which are impractical to be analysed one by one. We would like to help users prioritize, by highlighting insights that occurs more (or less) times than others, or that group a higher number of alerts, or alerts from different network types.

In this particular client, the NOC has a clear view of which parts of the network can benefit faster with AIOps insights, but since there is no easy way to search them to review in the UI, they are not inclined to use those insights in production. The idea with this enhancement request is to provide us a way to export data that seems to exist inside AIOps so users can perform macro analysis to find the most promising insights within the training results, and then inspect them further, and select those they are confident to enable in production.

For example: the UI has a page with data to justify the provided insight. It contains statistics like how often the pattern has been identified by the analysis, and how many and which alerts were present in each occurrence. If we were able to export this data into a spreadsheet, or at least a JSON file for each insight, we could help users refine their filters, and feel more inclined to enable the generated insights.

We would really appreciate to be able to export this data and work on it localy to avoid latency, and eventual limitations the UI may need to enforce so it remains responsive with a large amount of data. But being able to visualize this data in the UI is also a good alternative, if it provides flexible filter options.