ENH: Add better confusion matrix plot

pandas-ml / pandas-ml

pandas, scikit-learn, xgboost and seaborn integration

BSD 3-Clause "New" or "Revised" License

318 stars 78 forks source link

ENH: Add better confusion matrix plot #97

Open sinhrks opened 8 years ago

sinhrks commented 8 years ago

closes #92, closes #93, closes #94.

discretize the colorbar and adjust label loc
add annotate keyword to show annotations
remove backend keyword to output single & consistent plot

@scls19fr If it looks ok, I'm adding some tests.

cm = pdml.ConfusionMatrix([1, 2, 1, 1], [1, 2, 1, 2])
cm.plot(figsize=(5, 5))

index2

cm = pdml.ConfusionMatrix(np.random.choice([1, 2, 3, 4], 100), np.random.choice([1, 2, 3, 4], 100))
cm.plot(figsize=(5, 5), annotate=True)

index

coveralls commented 8 years ago

Coverage decreased (-0.003%) to 94.778% when pulling 18907db0336a777604afe03e6ce9dd8eb44105d7 on sinhrks:confmat_plot into 56c36199f37fa461def618d6050dde99bde63a46 on pandas-ml:master.

s-celles commented 8 years ago

That's very nice.

capture d ecran 2016-11-08 a 06 16 16

Does ticks can be moved to be at boundaries ?

How a normalized CM is displayed ?