Features

This spring brought an excellent crop of Haiku.

The winner is

voting and stacking,
bagging, boosting, random forests
error below zero

Grigorios Tsoumakas, Greece

(This highlights both advanced algorithms, and the dangers of overfitting, and achieves a zen quality with the "error below zero" reference - a data mining equivalent of "what is the sound of one hand clapping").

Honorable mentions include:

Model nearly done
Wait! Data from the future
It predicts too well

no deviations
subjective quality norm
sense of humor good

I was like her
according to Pearson R
we were both outliers

Here is also a Tanka, a longer version (with lines of 5-7-5-7-7 syllables)

Before K.D.D.
Data were data, pristine.
Hypotheses pure.
During, white water frenzy.
After, peaceful serene thoughts.

Ross Bettinger, USA