Weighting method documented in ML Model Results?

[Marlan]

Hi all,

I have been unable to find documentation of the weighting method setting in the model results summary.  Is it not there or am I just somehow missing it? 

I typically compare the performance of weighting (class weights) and no weighting methods when tuning my models. I'd like to be able to look at the results summary and see whether weighting was used.

Thanks,

Marlan

---

Operating system used: RHEL

[AdrienL · Accepted solution]

Hi,

There is no fully explicit reminder of the exact weighting method, you'd have to use the API to retrieve that. We will consider this feedback to add this explicitly on the UI in a future version of DSS.

If it can help, there is an indication of whether sample weights were used (i.e. if either the "Sample weights" or the "Class and sample weights" method was selected). If so:

• In the Features tab, the corresponding feature shows up as Weights
• In the Metrics tab, all metrics appear as "Weighted [metric name]"
• In the Summary tab and the model snippet (in the session view), there is a Sample weights variable mention
• In multiple result tabs, there is a ⚖️ icon on top with the name of the weights column + a tooltip on hover

[Marlan]

Thanks much for the reply @AdrienL. 

Good to know about the indications of whether Sample weights or Class and sample weights were used. I'm typically switching between Class weights and no weighting though. 

Yes, it would be great to have the weighting method documented. Everything else is so it seems reasonable that this would be too so there is complete documentation.

Thanks,

Marlan