Github user holdenk commented on a diff in the pull request:
https://github.com/apache/spark/pull/17494#discussion_r110127639
--- Diff: python/pyspark/ml/stat.py ---
@@ -71,6 +71,67 @@ def test(dataset, featuresCol, labelCol):
return _java2py(sc, javaTestObj.test(*args))
+class Correlation(object):
+ """
+ .. note:: Experimental
+
+ Compute the correlation matrix for the input dataset of Vectors using
the specified method.
+ Methods currently supported: `pearson` (default), `spearman`.
+
+ .. note:: For Spearman, a rank correlation, we need to create an
RDD[Double] for each column
+ and sort it in order to retrieve the ranks and then join the columns
back into an RDD[Vector],
+ which is fairly costly. Cache the input Dataset before calling corr
with `method = 'spearman'`
+ to avoid recomputing the common lineage.
+
+ :param dataset:
+ A dataset or a dataframe.
+ :param column:
+ The name of the column of vectors for which the correlation
coefficient needs
+ to be computed. This must be a column of the dataset, and it must
contain
+ Vector objects.
+ :param method:
+ String specifying the method to use for computing correlation.
+ Supported: `pearson` (default), `spearman`.
+ :return:
+ A dataframe that contains the correlation matrix of the column of
vectors. This
+ dataframe contains a single row and a single column of name
+ '$METHODNAME($COLUMN)'.
+
+ >>> from pyspark.ml.linalg import Vectors
+ >>> from pyspark.ml.stat import Correlation
+ >>> dataset = [[Vectors.dense([1, 0, 0, -2])],
+ ... [Vectors.dense([4, 5, 0, 3])],
+ ... [Vectors.dense([6, 7, 0, 8])],
+ ... [Vectors.dense([9, 0, 0, 1])]]
+ >>> dataset = spark.createDataFrame(dataset, ['features'])
+ >>> pearsonCorr = Correlation.corr(dataset, 'features',
'pearson').collect()[0][0]
+ >>> print(str(pearsonCorr).replace('nan', 'NaN'))
+ DenseMatrix([[ 1. , 0.05564149, NaN, 0.40047142],
--- End diff --
So maybe I'm being overly cautious - but doctests with floats have bit me
in the past - would it be good to use the ... syntax here or is this going to
be ok? (Just asking).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
