[SPARK-15883][MLLIB][DOCS] Fix broken links in mllib documents

## What changes were proposed in this pull request?

This issue fixes all broken links on Spark 2.0 preview MLLib documents. Also, this contains some editorial change.

**Fix broken links**
  * mllib-data-types.md
  * mllib-decision-tree.md
  * mllib-ensembles.md
  * mllib-feature-extraction.md
  * mllib-pmml-model-export.md
  * mllib-statistics.md

**Fix malformed section header and scala coding style**
  * mllib-linear-methods.md

**Replace indirect forward links with direct one**
  * ml-classification-regression.md

## How was this patch tested?

Manual tests (with `cd docs; jekyll build`.)

Author: Dongjoon Hyun <dongjoon@apache.org>

Closes #13608 from dongjoon-hyun/SPARK-15883.

1 files changed, 2 insertions, 2 deletions

diff --git a/docs/ml-classification-regression.md b/docs/ml-classification-regression.md
index 88457d4bb1..d7e5521cbc 100644
--- a/docs/ml-classification-regression.md
+++ b/docs/ml-classification-regression.md
@@ -815,7 +815,7 @@ The main differences between this API and the [original MLlib ensembles API](mll
 ## Random Forests
 
 [Random forests](http://en.wikipedia.org/wiki/Random_forest)
-are ensembles of [decision trees](ml-decision-tree.html).
+are ensembles of [decision trees](ml-classification-regression.html#decision-trees).
 Random forests combine many decision trees in order to reduce the risk of overfitting.
 The `spark.ml` implementation supports random forests for binary and multiclass classification and for regression,
 using both continuous and categorical features.
@@ -896,7 +896,7 @@ All output columns are optional; to exclude an output column, set its correspond
 ## Gradient-Boosted Trees (GBTs)
 
 [Gradient-Boosted Trees (GBTs)](http://en.wikipedia.org/wiki/Gradient_boosting)
-are ensembles of [decision trees](ml-decision-tree.html).
+are ensembles of [decision trees](ml-classification-regression.html#decision-trees).
 GBTs iteratively train decision trees in order to minimize a loss function.
 The `spark.ml` implementation supports GBTs for binary classification and for regression,
 using both continuous and categorical features.