Add alpha support for scatter plots

Describe the solution you’d like When I’m plotting scatter plots (such as PCA), oftentimes there is a bunch of data. By using the alpha parameter on matplotlib (also available through the pandas plotting interface), it allows you to discern the data and amount of overlapping.

Is your feature request related to a problem? Please describe. In general, I think an alpha parameter to any scatter plot visualizations (PCADecomposition) would be awesome. It would also be inline with what pandas plotting provides.

Examples Seaborn scatter plots come with a slight alpha applied. But you can use scatter_kws={'alpha':0.3} to set it. In pandas or matplotlib you just pass alpha=0.3.

Jake’s book also has alpha turned on when plotting scatter plots. See https://jakevdp.github.io/PythonDataScienceHandbook/05.09-principal-component-analysis.html

[mattharrison]

For alpha= vs scatter_kws, I think it depends on whether you want to tie the implementation to matplotlib. The former is generic, the latter indicates (to me at least) that these are the parameters that will be passed into matplotlib.

[bbengfort]

#554 adds alpha to the yellowbrick.contrib.ScatterVisualizer. Need to:

• open issue for scatter plot mixin
• open issue for non alpha legend mechanism
• open issue for other scatter visualizers that could use alpha params.

Then we can close this issue.