dirty-data-science / python

Tutorial material on machine learning with dirty data in Python

Machine-learning on dirty-data in Python: a tutorial

python's Issues

54e09fbf2f0bfe59c9d221ace8b136a7516b6ba2mads

reproducing notebook 1 with my data

Hi,
thanks a bunch for these two super valuable blog posts. I learned a good portion while reading them.

I am currently working on dirty data, a table of 224 rows and 27 columns. I created a clean subset and was playing with HistGradientBoostingRegressor.
I am not sure I understand this regressor yet completely, but I observe that it uses a half_least_squares loss by default, so I'd expect that scores are always positive.

If I run similar code than yours, I get the following:

In [41]:
hgbr = HistGradientBoostingRegressor()
score_hgbr = cross_val_score(hgbr, clean_X, clean_y, cv=5)
score_hgbr
Out [43]:
array([-0.26472938, -0.1909843 ,  0.24062092, -1.65719074, -0.40523002])

I am a bit irritated by the negative numbers in the scoring output. Any ideas where they may come from?