[Machine Learning Export] - Ability to share Machine Learning data




Is there a functionality to share the generated results from machine learning (either exporting the data or the Kibana UI) or any related posts to this topic?


(Jen Huang) #2

Hi, it's not possible to export the ML graphs, but you can retrieve the results programmatically through the Machine Learning Results APIs: https://www.elastic.co/guide/en/elasticsearch/reference/6.0/ml-apis.html#ml-api-result-endpoint

The results are returned in JSON. With further processing, the results could be converted to csv or another format for sharing purposes.


Hi Jen,

Thank you for the information, I was able to extract some of the ML data. However, it seems to only be grabbing anomalies only when I am posting a GET request for records. Is this the current feature for the ML data export?


(rich collier) #4

Hi Eric,

What data are you looking for?


Hi Rich,

I am new to this stack and was hoping to export all the data points that were used when graphing the ML job so I could generate the exact graph from the GET requests.

In addition, my GET requests return, for example, 90 records according to the count, but on the ML Kibana, I don't see as many listed in the anomalies. Do you know why that is?


(rich collier) #6


The first thing to know is that the index pattern that contains the information is called .ml-anomalies-* (there are index names like .ml-anomalies-shared, .ml-anomalies-jobname, etc.)

The next thing is that in those indices, there is always a key:value pair for the job_id. So, for example, job_id:orders

Lastly, in the indices, there can be 5 main kinds of results (keyed with a field called result_type)

  1. bucket
  2. record
  3. influencer
  4. model_forecast
  5. model_plot

(there are a few other types, but not worth talking about at the moment)

Each kind of result_type contains different information, but it is a summary of information gleaned from the analysis of the original raw documents that ML analyzed.

When you say "the data points that were used when graphing the ML job" - which graph are you thinking of? (perhaps also show a screenshot?)

(system) closed #7

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.