whizzml / examples Goto Github PK

Hi there,

Thanks a lot for building this awesome library and script! I was able to run it on my BigML datasets and it produced tons of good stuff in minutes.

Upon examining the shortlisted features, I've noticed "score" that happens to be superior to all other features left in the Fusion. I sense this came from one of the unsupervised models but I'm struggling to understand the exact whereabouts.

Can someone please advise how this "score" is calculated and what would it mean that it has the largest weight among all the selected features in my Fusion (Two bootstrap decision forests and one deepnet)? Looking forward to some guidance.

When running the SMACdown ensemble (https://github.com/whizzml/examples/blob/master/smacdown-ensemble/script.whizzml), there is an option to specify the "metric". The description text for this input states "Evaluation metric that we want to optimize. It must be a number". The default value is "average_phi". It seems if I enter "average_recall" this runs. It seems that the metric is used here:
33 phi (lambda (ev)
34 (let (metric-value (ev ["result" "model" metric] false))
Should the input text for "metric" be updated to be "Choose from average_recall, average_phi, accuracy, average_precision, or average_f_measure"?