[SPOT-173] [ML] training the model with history data/test data and use it to perform score calculations for real time data

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Labels: new-feature spot-ml
• Assignee: Unassigned
• Reporter: Anil
• Created: 03/Jun/17 16:56
• Updated: 05/Jun/17 15:42

Description

I want to have a kind of setup where I want to build a model with training data and use that model to generate scores for the real time records as model is a kind of definition to calculate metrics/scored results for the real data. I would like achieve that with SPOT for netflow, dns and proxy data. Currently I see in SPOT ML model is created and trained for each day and calculate scored results for the same day using the model.

I would like to have a kind of training phase where we create and train the model with test data or real data for certain period and use it to calculate scored results to filter events based on the trained model for real time analysis