[SPARK-19155][ML] Make family case insensitive in GLM

[actuaryzhang]

What changes were proposed in this pull request?

This is a supplement to PR #16516 which did not make the value from getFamily case insensitive. Current tests of poisson/binomial glm with weight fail when specifying 'Poisson' or 'Binomial', because the calculation of dispersion and pValue checks the value of family retrieved from getFamily

model.getFamily == Binomial.name || model.getFamily == Poisson.name

How was this patch tested?

Update existing tests for 'Poisson' and 'Binomial'.

@yanboliang@felixcheung@imatiach-msft

[actuaryzhang]

I would prefer that getFamily returns lower case values directly, because using getFamily.toLowerCase can get very cumbersome and I use this a lot in another PR #16344. If we want to keep getFamily to retrieve the raw value of family, then I can create a private method getFamilyLowerCase. Please advise.

[yanboliang]

@actuaryzhang I think the change is not appropriate, the function getFamily should return the raw value that users specified, this is the cause that I didn't change them in #16516 . Thanks.

[actuaryzhang]

@yanboliang Thanks for the quick response. How about the new commit, where I just change the value from getFamily to lower case when necessary, i.e., in the calculation of p-value and dispersion?

[yanboliang]

Looks good, I'll merge if it passes test. Thanks.

[SparkQA]

Test build #71819 has finished for PR 16675 at commit c2b4132.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #71820 has finished for PR 16675 at commit 97b0a1c.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[actuaryzhang]

@yanboliang Thanks. Seems to have passed tests.

[yanboliang]

LGTM, merged into master and branch-2.1. Thanks.