From 2adebe8de3881509e510fc518c562d1141ccd0ef Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Wed, 10 Aug 2016 13:40:18 +0800
Subject: [PATCH 1/7] add a chiSquare Selector based on False Positive Rate
 (FPR) test

---
 .../spark/mllib/feature/ChiSqSelector.scala   | 29 +++++++++++++++++--
 .../mllib/feature/ChiSqSelectorSuite.scala    | 18 ++++++++++++
 2 files changed, 45 insertions(+), 2 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index c8c2823bbaf04..f3316eeee8fec 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -173,8 +173,8 @@ object ChiSqSelectorModel extends Loader[ChiSqSelectorModel] {
  * Creates a ChiSquared feature selector.
  * @param numTopFeatures number of features that selector will select
  *                       (ordered by statistic value descending)
- *                       Note that if the number of features is < numTopFeatures, then this will
- *                       select all features.
+ *                       Note that if the number of features is less than numTopFeatures,
+ *                       then this will select all features.
  */
 @Since("1.3.0")
 class ChiSqSelector @Since("1.3.0") (
@@ -197,3 +197,28 @@ class ChiSqSelector @Since("1.3.0") (
     new ChiSqSelectorModel(indices)
   }
 }
+
+/**
+ * Creates a ChiSquared feature selector by False Positive Rate (FPR) test.
+ * @param alpha the highest p-value for features to be kept
+ */
+@Since("2.1.0")
+class ChiSqSelectorByFpr @Since("2.1.0") (
+  @Since("2.1.0") val alpha: Double) extends Serializable {
+
+  /**
+   * Returns a ChiSquared feature selector by FPR.
+   *
+   * @param data an `RDD[LabeledPoint]` containing the labeled dataset with categorical features.
+   *             Real-valued features will be treated as categorical for each distinct value.
+   *             Apply feature discretizer before using this function.
+   */
+  @Since("2.1.0")
+  def fit(data: RDD[LabeledPoint]): ChiSqSelectorModel = {
+    val indices = Statistics.chiSqTest(data)
+      .zipWithIndex.filter { case (res, _) => res.pValue < alpha }
+      .map { case (_, indices) => indices }
+      .sorted
+    new ChiSqSelectorModel(indices)
+  }
+}
diff --git a/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
index 734800a9afad6..6b2209c8a7c15 100644
--- a/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
@@ -65,6 +65,24 @@ class ChiSqSelectorSuite extends SparkFunSuite with MLlibTestSparkContext {
     assert(filteredData == preFilteredData)
   }
 
+  test("ChiSqSelectorByFpr transform test (sparse & dense vector)") {
+    val labeledDiscreteData = sc.parallelize(
+      Seq(LabeledPoint(0.0, Vectors.sparse(4, Array((0, 8.0), (1, 7.0)))),
+        LabeledPoint(1.0, Vectors.sparse(4, Array((1, 9.0), (2, 6.0), (3, 4.0)))),
+        LabeledPoint(1.0, Vectors.dense(Array(0.0, 9.0, 8.0, 4.0))),
+        LabeledPoint(2.0, Vectors.dense(Array(8.0, 9.0, 5.0, 9.0)))), 2)
+    val preFilteredData =
+      Set(LabeledPoint(0.0, Vectors.dense(Array(0.0))),
+        LabeledPoint(1.0, Vectors.dense(Array(4.0))),
+        LabeledPoint(1.0, Vectors.dense(Array(4.0))),
+        LabeledPoint(2.0, Vectors.dense(Array(9.0))))
+    val model = new ChiSqSelectorByFpr(0.1).fit(labeledDiscreteData)
+    val filteredData = labeledDiscreteData.map { lp =>
+      LabeledPoint(lp.label, model.transform(lp.features))
+    }.collect().toSet
+    assert(filteredData == preFilteredData)
+  }
+
   test("model load / save") {
     val model = ChiSqSelectorSuite.createModel()
     val tempDir = Utils.createTempDir()

From 7623563884355a04867ce5271baa286f65180e62 Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Tue, 16 Aug 2016 21:36:11 +0800
Subject: [PATCH 2/7] Configure the ChiSqSelector to reuse ChiSqTestResult by
 numTopFeatures, Percentile, and Fpr selector

---
 .../mllib/JavaChiSqSelectorExample.java       |   3 +-
 .../examples/mllib/ChiSqSelectorExample.scala |   3 +-
 .../spark/ml/feature/ChiSqSelector.scala      |  60 ++++++++++-
 .../mllib/api/python/PythonMLLibAPI.scala     |   4 +-
 .../spark/mllib/feature/ChiSqSelector.scala   | 102 +++++++++++-------
 .../mllib/feature/ChiSqSelectorSuite.scala    |   6 +-
 6 files changed, 126 insertions(+), 52 deletions(-)

diff --git a/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java b/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java
index ad44acb4cd6e3..f0619b7bc5685 100644
--- a/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java
+++ b/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java
@@ -56,7 +56,8 @@ public LabeledPoint call(LabeledPoint lp) {
     );
 
     // Create ChiSqSelector that will select top 50 of 692 features
-    ChiSqSelector selector = new ChiSqSelector(50);
+    ChiSqSelector selector = new ChiSqSelector();
+    selector.setNumTopFeatures(50);
     // Create ChiSqSelector model (selecting features)
     final ChiSqSelectorModel transformer = selector.fit(discretizedData.rdd());
     // Filter the top 50 features from each feature vector
diff --git a/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala b/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala
index 5e400b7d715b4..9fb520ce56acc 100644
--- a/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala
+++ b/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala
@@ -43,7 +43,8 @@ object ChiSqSelectorExample {
       LabeledPoint(lp.label, Vectors.dense(lp.features.toArray.map { x => (x / 16).floor }))
     }
     // Create ChiSqSelector that will select top 50 of 692 features
-    val selector = new ChiSqSelector(50)
+    val selector = new ChiSqSelector()
+    selector.setNumTopFeatures(50)
     // Create ChiSqSelector model (selecting features)
     val transformer = selector.fit(discretizedData)
     // Filter the top 50 features from each feature vector
diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
index 1482eb3d1f7a6..439514bdb4a4c 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
@@ -27,6 +27,7 @@ import org.apache.spark.ml.param._
 import org.apache.spark.ml.param.shared._
 import org.apache.spark.ml.util._
 import org.apache.spark.mllib.feature
+import org.apache.spark.mllib.feature.SelectorType
 import org.apache.spark.mllib.linalg.{Vectors => OldVectors}
 import org.apache.spark.mllib.regression.{LabeledPoint => OldLabeledPoint}
 import org.apache.spark.rdd.RDD
@@ -51,11 +52,29 @@ private[feature] trait ChiSqSelectorParams extends Params
       " number of features is < numTopFeatures, then this will select all features.",
     ParamValidators.gtEq(1))
   setDefault(numTopFeatures -> 50)
+  final val percentile = new IntParam(this, "percentile",
+    "Percentile of features that selector will select, ordered by statistics value descending.",
+    ParamValidators.gtEq(0))
+  setDefault(percentile -> 10)
+
+  final val alpha = new DoubleParam(this, "alpha",
+    "The highest p-value for features to be kept.",
+    ParamValidators.gtEq(0))
+  setDefault(alpha -> 0.05)
+
+  final val selectorType = SelectorType.KBest
 
   /** @group getParam */
   def getNumTopFeatures: Int = $(numTopFeatures)
+
+  def getPercentile: Int = $(percentile)
+
+  def getAlpha: Double = $(alpha)
+
+  def getSelectorType: SelectorType.Value = selectorType
 }
 
+
 /**
  * Chi-Squared feature selection, which selects categorical features to use for predicting a
  * categorical label.
@@ -66,10 +85,26 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
 
   @Since("1.6.0")
   def this() = this(Identifiable.randomUID("chiSqSelector"))
-
+  val chiSqSelector = new feature.ChiSqSelector()
   /** @group setParam */
   @Since("1.6.0")
-  def setNumTopFeatures(value: Int): this.type = set(numTopFeatures, value)
+  def setNumTopFeatures(value: Int): this.type = {
+    chiSqSelector.setNumTopFeatures(value)
+    chiSqSelector.setSelectorType(SelectorType.KBest)
+    set(numTopFeatures, value)
+  }
+
+  def setPercentile(value: Int): this.type = {
+    chiSqSelector.setPercentile(value)
+    chiSqSelector.setSelectorType(SelectorType.Percentile)
+    set(percentile, value)
+  }
+
+  def setAlpha(value: Double): this.type = {
+    chiSqSelector.setAlpha(value)
+    chiSqSelector.setSelectorType(SelectorType.Fpr)
+    set(alpha, value)
+  }
 
   /** @group setParam */
   @Since("1.6.0")
@@ -89,10 +124,25 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
     val input: RDD[OldLabeledPoint] =
       dataset.select(col($(labelCol)).cast(DoubleType), col($(featuresCol))).rdd.map {
         case Row(label: Double, features: Vector) =>
-          OldLabeledPoint(label, OldVectors.fromML(features))
+        OldLabeledPoint(label, OldVectors.fromML(features))
       }
-    val chiSqSelector = new feature.ChiSqSelector($(numTopFeatures)).fit(input)
-    copyValues(new ChiSqSelectorModel(uid, chiSqSelector).setParent(this))
+    val model = chiSqSelector.fit(input)
+    copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
+  }
+
+  def selectKBest(value: Int): ChiSqSelectorModel = {
+    val model = chiSqSelector.selectKBest(value)
+    copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
+  }
+
+  def selectPercentile(value: Int): ChiSqSelectorModel = {
+    val model = chiSqSelector.selectPercentile(value)
+    copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
+  }
+
+  def selectFpr(value: Double): ChiSqSelectorModel = {
+    val model = chiSqSelector.selectFpr(value)
+    copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
   @Since("1.6.0")
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala
index a80cca70f4b28..bdcfe70651e3d 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala
@@ -634,8 +634,8 @@ private[python] class PythonMLLibAPI extends Serializable {
    * Extra care needs to be taken in the Python code to ensure it gets freed on
    * exit; see the Py4J documentation.
    */
-  def fitChiSqSelector(numTopFeatures: Int, data: JavaRDD[LabeledPoint]): ChiSqSelectorModel = {
-    new ChiSqSelector(numTopFeatures).fit(data.rdd)
+  def fitChiSqSelector(data: JavaRDD[LabeledPoint]): ChiSqSelectorModel = {
+    new ChiSqSelector().fit(data.rdd)
   }
 
   /**
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index f3316eeee8fec..9bc75c65165fd 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -27,22 +27,26 @@ import org.apache.spark.annotation.Since
 import org.apache.spark.mllib.linalg.{DenseVector, SparseVector, Vector, Vectors}
 import org.apache.spark.mllib.regression.LabeledPoint
 import org.apache.spark.mllib.stat.Statistics
+import org.apache.spark.mllib.stat.test.ChiSqTestResult
 import org.apache.spark.mllib.util.{Loader, Saveable}
 import org.apache.spark.rdd.RDD
 import org.apache.spark.SparkContext
 import org.apache.spark.sql.{Row, SparkSession}
 
+object SelectorType extends Enumeration {
+  type SelectorType = Value
+  val KBest, Percentile, Fpr = Value
+}
+
 /**
  * Chi Squared selector model.
  *
- * @param selectedFeatures list of indices to select (filter). Must be ordered asc
+ * @param selectedFeatures list of indices to select (filter).
  */
 @Since("1.3.0")
 class ChiSqSelectorModel @Since("1.3.0") (
   @Since("1.3.0") val selectedFeatures: Array[Int]) extends VectorTransformer with Saveable {
 
-  require(isSorted(selectedFeatures), "Array has to be sorted asc")
-
   protected def isSorted(array: Array[Int]): Boolean = {
     var i = 1
     val len = array.length
@@ -69,21 +73,23 @@ class ChiSqSelectorModel @Since("1.3.0") (
    * Preserves the order of filtered features the same as their indices are stored.
    * Might be moved to Vector as .slice
    * @param features vector
-   * @param filterIndices indices of features to filter, must be ordered asc
+   * @param filterIndices indices of features to filter
    */
   private def compress(features: Vector, filterIndices: Array[Int]): Vector = {
+    val orderedIndices = filterIndices.sorted
+    require(isSorted(orderedIndices), "Array has to be sorted asc")
     features match {
       case SparseVector(size, indices, values) =>
-        val newSize = filterIndices.length
+        val newSize = orderedIndices.length
         val newValues = new ArrayBuilder.ofDouble
         val newIndices = new ArrayBuilder.ofInt
         var i = 0
         var j = 0
         var indicesIdx = 0
         var filterIndicesIdx = 0
-        while (i < indices.length && j < filterIndices.length) {
+        while (i < indices.length && j < orderedIndices.length) {
           indicesIdx = indices(i)
-          filterIndicesIdx = filterIndices(j)
+          filterIndicesIdx = orderedIndices(j)
           if (indicesIdx == filterIndicesIdx) {
             newIndices += j
             newValues += values(i)
@@ -101,7 +107,7 @@ class ChiSqSelectorModel @Since("1.3.0") (
         Vectors.sparse(newSize, newIndices.result(), newValues.result())
       case DenseVector(values) =>
         val values = features.toArray
-        Vectors.dense(filterIndices.map(i => values(i)))
+        Vectors.dense(orderedIndices.map(i => values(i)))
       case other =>
         throw new UnsupportedOperationException(
           s"Only sparse and dense vectors are supported but got ${other.getClass}.")
@@ -171,14 +177,34 @@ object ChiSqSelectorModel extends Loader[ChiSqSelectorModel] {
 
 /**
  * Creates a ChiSquared feature selector.
- * @param numTopFeatures number of features that selector will select
- *                       (ordered by statistic value descending)
- *                       Note that if the number of features is less than numTopFeatures,
- *                       then this will select all features.
  */
 @Since("1.3.0")
-class ChiSqSelector @Since("1.3.0") (
-  @Since("1.3.0") val numTopFeatures: Int) extends Serializable {
+class ChiSqSelector @Since("1.3.0") () extends Serializable {
+  var numTopFeatures: Int = 1
+  var percentile: Int = 10
+  var alpha: Double = 0.05
+  var selectorType = SelectorType.KBest
+  var chiSqTestResult: Array[ChiSqTestResult] = new Array[ChiSqTestResult](0)
+
+  def setNumTopFeatures(value: Int): this.type = {
+    numTopFeatures = value
+    selectorType = SelectorType.KBest
+    this
+  }
+  def setPercentile(value: Int): this.type = {
+    percentile = value
+    selectorType = SelectorType.Percentile
+    this
+  }
+  def setAlpha(value: Double): this.type = {
+    alpha = value
+    selectorType = SelectorType.Fpr
+    this
+  }
+  def setSelectorType(value: SelectorType.Value): this.type = {
+    selectorType = value
+    this
+  }
 
   /**
    * Returns a ChiSquared feature selector.
@@ -189,36 +215,32 @@ class ChiSqSelector @Since("1.3.0") (
    */
   @Since("1.3.0")
   def fit(data: RDD[LabeledPoint]): ChiSqSelectorModel = {
-    val indices = Statistics.chiSqTest(data)
-      .zipWithIndex.sortBy { case (res, _) => -res.statistic }
-      .take(numTopFeatures)
-      .map { case (_, indices) => indices }
-      .sorted
+    chiSqTestResult = Statistics.chiSqTest(data)
+      selectorType match {
+        case SelectorType.KBest => selectKBest(numTopFeatures)
+        case SelectorType.Percentile => selectPercentile(percentile)
+        case SelectorType.Fpr => selectFpr(alpha)
+      }
+  }
+
+  def selectKBest(value: Int): ChiSqSelectorModel = {
+    val indices = chiSqTestResult.zipWithIndex.sortBy { case (res, _) => -res.statistic }
+    .take(numTopFeatures)
+    .map { case (_, indices) => indices }
     new ChiSqSelectorModel(indices)
   }
-}
 
-/**
- * Creates a ChiSquared feature selector by False Positive Rate (FPR) test.
- * @param alpha the highest p-value for features to be kept
- */
-@Since("2.1.0")
-class ChiSqSelectorByFpr @Since("2.1.0") (
-  @Since("2.1.0") val alpha: Double) extends Serializable {
+  def selectPercentile(value: Int): ChiSqSelectorModel = {
+    val indices = chiSqTestResult.zipWithIndex.sortBy { case (res, _) => -res.statistic }
+    .take((chiSqTestResult.length * percentile / 100).toInt)
+    .map { case (_, indices) => indices }
+    new ChiSqSelectorModel(indices)
+  }
 
-  /**
-   * Returns a ChiSquared feature selector by FPR.
-   *
-   * @param data an `RDD[LabeledPoint]` containing the labeled dataset with categorical features.
-   *             Real-valued features will be treated as categorical for each distinct value.
-   *             Apply feature discretizer before using this function.
-   */
-  @Since("2.1.0")
-  def fit(data: RDD[LabeledPoint]): ChiSqSelectorModel = {
-    val indices = Statistics.chiSqTest(data)
-      .zipWithIndex.filter { case (res, _) => res.pValue < alpha }
-      .map { case (_, indices) => indices }
-      .sorted
+  def selectFpr(value: Double): ChiSqSelectorModel = {
+    val indices = chiSqTestResult.zipWithIndex.filter{ case (res, _) => res.pValue < alpha }
+    .map { case (_, indices) => indices }
     new ChiSqSelectorModel(indices)
   }
 }
+
diff --git a/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
index 6b2209c8a7c15..d61888df9c0dc 100644
--- a/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
@@ -58,14 +58,14 @@ class ChiSqSelectorSuite extends SparkFunSuite with MLlibTestSparkContext {
         LabeledPoint(1.0, Vectors.dense(Array(6.0))),
         LabeledPoint(1.0, Vectors.dense(Array(8.0))),
         LabeledPoint(2.0, Vectors.dense(Array(5.0))))
-    val model = new ChiSqSelector(1).fit(labeledDiscreteData)
+    val model = new ChiSqSelector().fit(labeledDiscreteData)
     val filteredData = labeledDiscreteData.map { lp =>
       LabeledPoint(lp.label, model.transform(lp.features))
     }.collect().toSet
     assert(filteredData == preFilteredData)
   }
 
-  test("ChiSqSelectorByFpr transform test (sparse & dense vector)") {
+  test("ChiSqSelector by FPR transform test (sparse & dense vector)") {
     val labeledDiscreteData = sc.parallelize(
       Seq(LabeledPoint(0.0, Vectors.sparse(4, Array((0, 8.0), (1, 7.0)))),
         LabeledPoint(1.0, Vectors.sparse(4, Array((1, 9.0), (2, 6.0), (3, 4.0)))),
@@ -76,7 +76,7 @@ class ChiSqSelectorSuite extends SparkFunSuite with MLlibTestSparkContext {
         LabeledPoint(1.0, Vectors.dense(Array(4.0))),
         LabeledPoint(1.0, Vectors.dense(Array(4.0))),
         LabeledPoint(2.0, Vectors.dense(Array(9.0))))
-    val model = new ChiSqSelectorByFpr(0.1).fit(labeledDiscreteData)
+    val model = new ChiSqSelector().setAlpha(0.1).fit(labeledDiscreteData)
     val filteredData = labeledDiscreteData.map { lp =>
       LabeledPoint(lp.label, model.transform(lp.features))
     }.collect().toSet

From 3d6aecb8441503c9c3d62a2d8a3d48824b9d6637 Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Wed, 17 Aug 2016 10:34:59 +0800
Subject: [PATCH 3/7] Config the ChiSqSelector to reuse the ChiSqTestResult by
 KBest, Percentile and FPR selector

---
 .../mllib/JavaChiSqSelectorExample.java       |  3 +-
 .../examples/mllib/ChiSqSelectorExample.scala |  3 +-
 .../spark/ml/feature/ChiSqSelector.scala      | 12 +++---
 .../mllib/api/python/PythonMLLibAPI.scala     |  4 +-
 .../spark/mllib/feature/ChiSqSelector.scala   | 38 ++++++++++---------
 .../mllib/feature/ChiSqSelectorSuite.scala    |  2 +-
 6 files changed, 32 insertions(+), 30 deletions(-)

diff --git a/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java b/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java
index f0619b7bc5685..ad44acb4cd6e3 100644
--- a/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java
+++ b/examples/src/main/java/org/apache/spark/examples/mllib/JavaChiSqSelectorExample.java
@@ -56,8 +56,7 @@ public LabeledPoint call(LabeledPoint lp) {
     );
 
     // Create ChiSqSelector that will select top 50 of 692 features
-    ChiSqSelector selector = new ChiSqSelector();
-    selector.setNumTopFeatures(50);
+    ChiSqSelector selector = new ChiSqSelector(50);
     // Create ChiSqSelector model (selecting features)
     final ChiSqSelectorModel transformer = selector.fit(discretizedData.rdd());
     // Filter the top 50 features from each feature vector
diff --git a/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala b/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala
index 9fb520ce56acc..5e400b7d715b4 100644
--- a/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala
+++ b/examples/src/main/scala/org/apache/spark/examples/mllib/ChiSqSelectorExample.scala
@@ -43,8 +43,7 @@ object ChiSqSelectorExample {
       LabeledPoint(lp.label, Vectors.dense(lp.features.toArray.map { x => (x / 16).floor }))
     }
     // Create ChiSqSelector that will select top 50 of 692 features
-    val selector = new ChiSqSelector()
-    selector.setNumTopFeatures(50)
+    val selector = new ChiSqSelector(50)
     // Create ChiSqSelector model (selecting features)
     val transformer = selector.fit(discretizedData)
     // Filter the top 50 features from each feature vector
diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
index 439514bdb4a4c..a44ac2fe73aea 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
@@ -27,7 +27,7 @@ import org.apache.spark.ml.param._
 import org.apache.spark.ml.param.shared._
 import org.apache.spark.ml.util._
 import org.apache.spark.mllib.feature
-import org.apache.spark.mllib.feature.SelectorType
+import org.apache.spark.mllib.feature.ChiSqSelectorType
 import org.apache.spark.mllib.linalg.{Vectors => OldVectors}
 import org.apache.spark.mllib.regression.{LabeledPoint => OldLabeledPoint}
 import org.apache.spark.rdd.RDD
@@ -62,7 +62,7 @@ private[feature] trait ChiSqSelectorParams extends Params
     ParamValidators.gtEq(0))
   setDefault(alpha -> 0.05)
 
-  final val selectorType = SelectorType.KBest
+  final val selectorType = ChiSqSelectorType.KBest
 
   /** @group getParam */
   def getNumTopFeatures: Int = $(numTopFeatures)
@@ -71,7 +71,7 @@ private[feature] trait ChiSqSelectorParams extends Params
 
   def getAlpha: Double = $(alpha)
 
-  def getSelectorType: SelectorType.Value = selectorType
+  def getChiSqSelectorType: ChiSqSelectorType.Value = selectorType
 }
 
 
@@ -90,19 +90,19 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
   @Since("1.6.0")
   def setNumTopFeatures(value: Int): this.type = {
     chiSqSelector.setNumTopFeatures(value)
-    chiSqSelector.setSelectorType(SelectorType.KBest)
+    chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.KBest)
     set(numTopFeatures, value)
   }
 
   def setPercentile(value: Int): this.type = {
     chiSqSelector.setPercentile(value)
-    chiSqSelector.setSelectorType(SelectorType.Percentile)
+    chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.Percentile)
     set(percentile, value)
   }
 
   def setAlpha(value: Double): this.type = {
     chiSqSelector.setAlpha(value)
-    chiSqSelector.setSelectorType(SelectorType.Fpr)
+    chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.Fpr)
     set(alpha, value)
   }
 
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala
index bdcfe70651e3d..a80cca70f4b28 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala
@@ -634,8 +634,8 @@ private[python] class PythonMLLibAPI extends Serializable {
    * Extra care needs to be taken in the Python code to ensure it gets freed on
    * exit; see the Py4J documentation.
    */
-  def fitChiSqSelector(data: JavaRDD[LabeledPoint]): ChiSqSelectorModel = {
-    new ChiSqSelector().fit(data.rdd)
+  def fitChiSqSelector(numTopFeatures: Int, data: JavaRDD[LabeledPoint]): ChiSqSelectorModel = {
+    new ChiSqSelector(numTopFeatures).fit(data.rdd)
   }
 
   /**
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index 9bc75c65165fd..e2345b85a279e 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -33,7 +33,7 @@ import org.apache.spark.rdd.RDD
 import org.apache.spark.SparkContext
 import org.apache.spark.sql.{Row, SparkSession}
 
-object SelectorType extends Enumeration {
+object ChiSqSelectorType extends Enumeration {
   type SelectorType = Value
   val KBest, Percentile, Fpr = Value
 }
@@ -77,7 +77,6 @@ class ChiSqSelectorModel @Since("1.3.0") (
    */
   private def compress(features: Vector, filterIndices: Array[Int]): Vector = {
     val orderedIndices = filterIndices.sorted
-    require(isSorted(orderedIndices), "Array has to be sorted asc")
     features match {
       case SparseVector(size, indices, values) =>
         val newSize = orderedIndices.length
@@ -178,30 +177,34 @@ object ChiSqSelectorModel extends Loader[ChiSqSelectorModel] {
 /**
  * Creates a ChiSquared feature selector.
  */
-@Since("1.3.0")
-class ChiSqSelector @Since("1.3.0") () extends Serializable {
-  var numTopFeatures: Int = 1
-  var percentile: Int = 10
-  var alpha: Double = 0.05
-  var selectorType = SelectorType.KBest
-  var chiSqTestResult: Array[ChiSqTestResult] = new Array[ChiSqTestResult](0)
-
+@Since("2.1.0")
+class ChiSqSelector @Since("2.1.0") () extends Serializable {
+  private var numTopFeatures: Int = 1
+  private var percentile: Int = 10
+  private var alpha: Double = 0.05
+  private var selectorType = ChiSqSelectorType.KBest
+  private var chiSqTestResult: Array[ChiSqTestResult] = _
+
+  def this(numTopFeatures: Int) {
+    this()
+    this.numTopFeatures = numTopFeatures
+  }
   def setNumTopFeatures(value: Int): this.type = {
     numTopFeatures = value
-    selectorType = SelectorType.KBest
+    selectorType = ChiSqSelectorType.KBest
     this
   }
   def setPercentile(value: Int): this.type = {
     percentile = value
-    selectorType = SelectorType.Percentile
+    selectorType = ChiSqSelectorType.Percentile
     this
   }
   def setAlpha(value: Double): this.type = {
     alpha = value
-    selectorType = SelectorType.Fpr
+    selectorType = ChiSqSelectorType.Fpr
     this
   }
-  def setSelectorType(value: SelectorType.Value): this.type = {
+  def setChiSqSelectorType(value: ChiSqSelectorType.Value): this.type = {
     selectorType = value
     this
   }
@@ -217,9 +220,10 @@ class ChiSqSelector @Since("1.3.0") () extends Serializable {
   def fit(data: RDD[LabeledPoint]): ChiSqSelectorModel = {
     chiSqTestResult = Statistics.chiSqTest(data)
       selectorType match {
-        case SelectorType.KBest => selectKBest(numTopFeatures)
-        case SelectorType.Percentile => selectPercentile(percentile)
-        case SelectorType.Fpr => selectFpr(alpha)
+        case ChiSqSelectorType.KBest => selectKBest(numTopFeatures)
+        case ChiSqSelectorType.Percentile => selectPercentile(percentile)
+        case ChiSqSelectorType.Fpr => selectFpr(alpha)
+        case _ => throw new Exception
       }
   }
 
diff --git a/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
index d61888df9c0dc..e181a544f7159 100644
--- a/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/mllib/feature/ChiSqSelectorSuite.scala
@@ -58,7 +58,7 @@ class ChiSqSelectorSuite extends SparkFunSuite with MLlibTestSparkContext {
         LabeledPoint(1.0, Vectors.dense(Array(6.0))),
         LabeledPoint(1.0, Vectors.dense(Array(8.0))),
         LabeledPoint(2.0, Vectors.dense(Array(5.0))))
-    val model = new ChiSqSelector().fit(labeledDiscreteData)
+    val model = new ChiSqSelector(1).fit(labeledDiscreteData)
     val filteredData = labeledDiscreteData.map { lp =>
       LabeledPoint(lp.label, model.transform(lp.features))
     }.collect().toSet

From b522c5a0d8cbd5a545396117ca60890853f8a6d9 Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Wed, 17 Aug 2016 17:59:16 +0800
Subject: [PATCH 4/7] add Since annotation

---
 .../spark/ml/feature/ChiSqSelector.scala      |  7 ++++-
 .../spark/mllib/feature/ChiSqSelector.scala   | 27 ++++++++++++++-----
 .../spark/ml/feature/ChiSqSelectorSuite.scala | 11 ++++++--
 3 files changed, 35 insertions(+), 10 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
index a44ac2fe73aea..5cfc6f036e0a1 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
@@ -87,19 +87,21 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
   def this() = this(Identifiable.randomUID("chiSqSelector"))
   val chiSqSelector = new feature.ChiSqSelector()
   /** @group setParam */
-  @Since("1.6.0")
+  @Since("2.1.0")
   def setNumTopFeatures(value: Int): this.type = {
     chiSqSelector.setNumTopFeatures(value)
     chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.KBest)
     set(numTopFeatures, value)
   }
 
+  @Since("2.1.0")
   def setPercentile(value: Int): this.type = {
     chiSqSelector.setPercentile(value)
     chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.Percentile)
     set(percentile, value)
   }
 
+  @Since("2.1.0")
   def setAlpha(value: Double): this.type = {
     chiSqSelector.setAlpha(value)
     chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.Fpr)
@@ -130,16 +132,19 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
+  @Since("2.1.0")
   def selectKBest(value: Int): ChiSqSelectorModel = {
     val model = chiSqSelector.selectKBest(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
+  @Since("2.1.0")
   def selectPercentile(value: Int): ChiSqSelectorModel = {
     val model = chiSqSelector.selectPercentile(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
+  @Since("2.1.0")
   def selectFpr(value: Double): ChiSqSelectorModel = {
     val model = chiSqSelector.selectFpr(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index e2345b85a279e..a5069d414dc5c 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -33,6 +33,7 @@ import org.apache.spark.rdd.RDD
 import org.apache.spark.SparkContext
 import org.apache.spark.sql.{Row, SparkSession}
 
+@Since("2.1.0")
 object ChiSqSelectorType extends Enumeration {
   type SelectorType = Value
   val KBest, Percentile, Fpr = Value
@@ -179,31 +180,40 @@ object ChiSqSelectorModel extends Loader[ChiSqSelectorModel] {
  */
 @Since("2.1.0")
 class ChiSqSelector @Since("2.1.0") () extends Serializable {
-  private var numTopFeatures: Int = 1
+  private var numTopFeatures: Int = 50
   private var percentile: Int = 10
   private var alpha: Double = 0.05
   private var selectorType = ChiSqSelectorType.KBest
   private var chiSqTestResult: Array[ChiSqTestResult] = _
 
+  @Since("1.3.0")
   def this(numTopFeatures: Int) {
     this()
     this.numTopFeatures = numTopFeatures
   }
+  
+  @Since("2.1.0")
   def setNumTopFeatures(value: Int): this.type = {
     numTopFeatures = value
     selectorType = ChiSqSelectorType.KBest
     this
   }
+  
+  @Since("2.1.0")
   def setPercentile(value: Int): this.type = {
     percentile = value
     selectorType = ChiSqSelectorType.Percentile
     this
   }
+  
+  @Since("2.1.0")
   def setAlpha(value: Double): this.type = {
     alpha = value
     selectorType = ChiSqSelectorType.Fpr
     this
   }
+  
+  @Since("2.1.0")
   def setChiSqSelectorType(value: ChiSqSelectorType.Value): this.type = {
     selectorType = value
     this
@@ -219,14 +229,15 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
   @Since("1.3.0")
   def fit(data: RDD[LabeledPoint]): ChiSqSelectorModel = {
     chiSqTestResult = Statistics.chiSqTest(data)
-      selectorType match {
-        case ChiSqSelectorType.KBest => selectKBest(numTopFeatures)
-        case ChiSqSelectorType.Percentile => selectPercentile(percentile)
-        case ChiSqSelectorType.Fpr => selectFpr(alpha)
-        case _ => throw new Exception
-      }
+    selectorType match {
+      case ChiSqSelectorType.KBest => selectKBest(numTopFeatures)
+      case ChiSqSelectorType.Percentile => selectPercentile(percentile)
+      case ChiSqSelectorType.Fpr => selectFpr(alpha)
+      case _ => throw new Exception("Unknown ChiSqSelector Type")
+    }
   }
 
+  @Since("2.1.0")
   def selectKBest(value: Int): ChiSqSelectorModel = {
     val indices = chiSqTestResult.zipWithIndex.sortBy { case (res, _) => -res.statistic }
     .take(numTopFeatures)
@@ -234,6 +245,7 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
     new ChiSqSelectorModel(indices)
   }
 
+  @Since("2.1.0")
   def selectPercentile(value: Int): ChiSqSelectorModel = {
     val indices = chiSqTestResult.zipWithIndex.sortBy { case (res, _) => -res.statistic }
     .take((chiSqTestResult.length * percentile / 100).toInt)
@@ -241,6 +253,7 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
     new ChiSqSelectorModel(indices)
   }
 
+  @Since("2.1.0")
   def selectFpr(value: Double): ChiSqSelectorModel = {
     val indices = chiSqTestResult.zipWithIndex.filter{ case (res, _) => res.pValue < alpha }
     .map { case (_, indices) => indices }
diff --git a/mllib/src/test/scala/org/apache/spark/ml/feature/ChiSqSelectorSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/feature/ChiSqSelectorSuite.scala
index 3558290b23ae0..a29ff83ae0cce 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/feature/ChiSqSelectorSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/feature/ChiSqSelectorSuite.scala
@@ -49,16 +49,23 @@ class ChiSqSelectorSuite extends SparkFunSuite with MLlibTestSparkContext
       .map(x => (x._1.label, x._1.features, x._2))
       .toDF("label", "data", "preFilteredData")
 
-    val model = new ChiSqSelector()
+    val selector = new ChiSqSelector()
       .setNumTopFeatures(1)
       .setFeaturesCol("data")
       .setLabelCol("label")
       .setOutputCol("filtered")
 
-    model.fit(df).transform(df).select("filtered", "preFilteredData").collect().foreach {
+    selector.fit(df).transform(df).select("filtered", "preFilteredData").collect().foreach {
       case Row(vec1: Vector, vec2: Vector) =>
         assert(vec1 ~== vec2 absTol 1e-1)
     }
+
+    selector.selectPercentile(34).transform(df)
+    .select("filtered", "preFilteredData").collect().foreach {
+      case Row(vec1: Vector, vec2: Vector) =>
+        assert(vec1 ~== vec2 absTol 1e-1)
+    }
+
   }
 
   test("ChiSqSelector read/write") {

From 3431a7a0dada65e06ab87d6cbac8fd36222c0a32 Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Thu, 18 Aug 2016 11:00:42 +0800
Subject: [PATCH 5/7] Fix the bugs of Save/Load ML/ChiSqSelector

---
 .../spark/ml/feature/ChiSqSelector.scala      | 43 +++++++++++++------
 .../spark/mllib/feature/ChiSqSelector.scala   |  8 ++--
 2 files changed, 33 insertions(+), 18 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
index 5cfc6f036e0a1..a936790bf1697 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
@@ -52,26 +52,32 @@ private[feature] trait ChiSqSelectorParams extends Params
       " number of features is < numTopFeatures, then this will select all features.",
     ParamValidators.gtEq(1))
   setDefault(numTopFeatures -> 50)
+
+  /** @group getParam */
+  def getNumTopFeatures: Int = $(numTopFeatures)
+
   final val percentile = new IntParam(this, "percentile",
     "Percentile of features that selector will select, ordered by statistics value descending.",
     ParamValidators.gtEq(0))
   setDefault(percentile -> 10)
 
+  /** @group getParam */
+  def getPercentile: Int = $(percentile)
+
   final val alpha = new DoubleParam(this, "alpha",
     "The highest p-value for features to be kept.",
     ParamValidators.gtEq(0))
   setDefault(alpha -> 0.05)
 
-  final val selectorType = ChiSqSelectorType.KBest
-
   /** @group getParam */
-  def getNumTopFeatures: Int = $(numTopFeatures)
-
-  def getPercentile: Int = $(percentile)
-
   def getAlpha: Double = $(alpha)
 
-  def getChiSqSelectorType: ChiSqSelectorType.Value = selectorType
+  final val selectorType = new Param[String](this, "selectorType",
+    "ChiSqSelector Type: KBest, Percentile, Fpr")
+  setDefault(selectorType -> ChiSqSelectorType.KBest.toString)
+
+  /** @group getParam */
+  def getChiSqSelectorType: String = $(selectorType)
 }
 
 
@@ -85,26 +91,23 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
 
   @Since("1.6.0")
   def this() = this(Identifiable.randomUID("chiSqSelector"))
-  val chiSqSelector = new feature.ChiSqSelector()
+  var chiSqSelector: feature.ChiSqSelector = null
   /** @group setParam */
   @Since("2.1.0")
   def setNumTopFeatures(value: Int): this.type = {
-    chiSqSelector.setNumTopFeatures(value)
-    chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.KBest)
+    set(selectorType, ChiSqSelectorType.KBest.toString)
     set(numTopFeatures, value)
   }
 
   @Since("2.1.0")
   def setPercentile(value: Int): this.type = {
-    chiSqSelector.setPercentile(value)
-    chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.Percentile)
+    set(selectorType, ChiSqSelectorType.Percentile.toString)
     set(percentile, value)
   }
 
   @Since("2.1.0")
   def setAlpha(value: Double): this.type = {
-    chiSqSelector.setAlpha(value)
-    chiSqSelector.setChiSqSelectorType(ChiSqSelectorType.Fpr)
+    set(selectorType, ChiSqSelectorType.Fpr.toString)
     set(alpha, value)
   }
 
@@ -128,24 +131,36 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
         case Row(label: Double, features: Vector) =>
         OldLabeledPoint(label, OldVectors.fromML(features))
       }
+    $(selectorType) match {
+      case "KBest" =>
+        chiSqSelector = new feature.ChiSqSelector().setNumTopFeatures($(numTopFeatures))
+      case "Percentile" =>
+        chiSqSelector = new feature.ChiSqSelector().setPercentile($(percentile))
+      case "Fpr" =>
+        chiSqSelector = new feature.ChiSqSelector().setAlpha($(alpha))
+      case _ => throw new Exception("Unknown ChiSqSelector Type.")
+    }
     val model = chiSqSelector.fit(input)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
   @Since("2.1.0")
   def selectKBest(value: Int): ChiSqSelectorModel = {
+    require(chiSqSelector != null, "ChiSqSelector has not been created.")
     val model = chiSqSelector.selectKBest(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
   @Since("2.1.0")
   def selectPercentile(value: Int): ChiSqSelectorModel = {
+    require(chiSqSelector != null, "ChiSqSelector has not been created.")
     val model = chiSqSelector.selectPercentile(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
 
   @Since("2.1.0")
   def selectFpr(value: Double): ChiSqSelectorModel = {
+    require(chiSqSelector != null, "ChiSqSelector has not been created.")
     val model = chiSqSelector.selectFpr(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
   }
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index a5069d414dc5c..b58e286560388 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -191,28 +191,28 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
     this()
     this.numTopFeatures = numTopFeatures
   }
-  
+
   @Since("2.1.0")
   def setNumTopFeatures(value: Int): this.type = {
     numTopFeatures = value
     selectorType = ChiSqSelectorType.KBest
     this
   }
-  
+
   @Since("2.1.0")
   def setPercentile(value: Int): this.type = {
     percentile = value
     selectorType = ChiSqSelectorType.Percentile
     this
   }
-  
+
   @Since("2.1.0")
   def setAlpha(value: Double): this.type = {
     alpha = value
     selectorType = ChiSqSelectorType.Fpr
     this
   }
-  
+
   @Since("2.1.0")
   def setChiSqSelectorType(value: ChiSqSelectorType.Value): this.type = {
     selectorType = value

From 89e2dd5df055d55fe627e679c9c790d592203c8d Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Thu, 18 Aug 2016 12:39:48 +0800
Subject: [PATCH 6/7] Change ChiSqSelectorType in ml to String

---
 .../spark/ml/feature/ChiSqSelector.scala      | 19 +++++++++----------
 .../spark/mllib/feature/ChiSqSelector.scala   |  6 +++---
 2 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
index a936790bf1697..5a53930135f18 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
@@ -27,7 +27,6 @@ import org.apache.spark.ml.param._
 import org.apache.spark.ml.param.shared._
 import org.apache.spark.ml.util._
 import org.apache.spark.mllib.feature
-import org.apache.spark.mllib.feature.ChiSqSelectorType
 import org.apache.spark.mllib.linalg.{Vectors => OldVectors}
 import org.apache.spark.mllib.regression.{LabeledPoint => OldLabeledPoint}
 import org.apache.spark.rdd.RDD
@@ -56,13 +55,13 @@ private[feature] trait ChiSqSelectorParams extends Params
   /** @group getParam */
   def getNumTopFeatures: Int = $(numTopFeatures)
 
-  final val percentile = new IntParam(this, "percentile",
+  final val percentile = new DoubleParam(this, "percentile",
     "Percentile of features that selector will select, ordered by statistics value descending.",
     ParamValidators.gtEq(0))
   setDefault(percentile -> 10)
 
   /** @group getParam */
-  def getPercentile: Int = $(percentile)
+  def getPercentile: Double = $(percentile)
 
   final val alpha = new DoubleParam(this, "alpha",
     "The highest p-value for features to be kept.",
@@ -74,7 +73,7 @@ private[feature] trait ChiSqSelectorParams extends Params
 
   final val selectorType = new Param[String](this, "selectorType",
     "ChiSqSelector Type: KBest, Percentile, Fpr")
-  setDefault(selectorType -> ChiSqSelectorType.KBest.toString)
+  setDefault(selectorType -> "KBest")
 
   /** @group getParam */
   def getChiSqSelectorType: String = $(selectorType)
@@ -95,19 +94,19 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
   /** @group setParam */
   @Since("2.1.0")
   def setNumTopFeatures(value: Int): this.type = {
-    set(selectorType, ChiSqSelectorType.KBest.toString)
+    set(selectorType, "KBest")
     set(numTopFeatures, value)
   }
 
   @Since("2.1.0")
-  def setPercentile(value: Int): this.type = {
-    set(selectorType, ChiSqSelectorType.Percentile.toString)
+  def setPercentile(value: Double): this.type = {
+    set(selectorType, "Percentile")
     set(percentile, value)
   }
 
   @Since("2.1.0")
   def setAlpha(value: Double): this.type = {
-    set(selectorType, ChiSqSelectorType.Fpr.toString)
+    set(selectorType, "Fpr")
     set(alpha, value)
   }
 
@@ -129,7 +128,7 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
     val input: RDD[OldLabeledPoint] =
       dataset.select(col($(labelCol)).cast(DoubleType), col($(featuresCol))).rdd.map {
         case Row(label: Double, features: Vector) =>
-        OldLabeledPoint(label, OldVectors.fromML(features))
+          OldLabeledPoint(label, OldVectors.fromML(features))
       }
     $(selectorType) match {
       case "KBest" =>
@@ -152,7 +151,7 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
   }
 
   @Since("2.1.0")
-  def selectPercentile(value: Int): ChiSqSelectorModel = {
+  def selectPercentile(value: Double): ChiSqSelectorModel = {
     require(chiSqSelector != null, "ChiSqSelector has not been created.")
     val model = chiSqSelector.selectPercentile(value)
     copyValues(new ChiSqSelectorModel(uid, model).setParent(this))
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index b58e286560388..eb4a24214efd5 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -181,7 +181,7 @@ object ChiSqSelectorModel extends Loader[ChiSqSelectorModel] {
 @Since("2.1.0")
 class ChiSqSelector @Since("2.1.0") () extends Serializable {
   private var numTopFeatures: Int = 50
-  private var percentile: Int = 10
+  private var percentile: Double = 10
   private var alpha: Double = 0.05
   private var selectorType = ChiSqSelectorType.KBest
   private var chiSqTestResult: Array[ChiSqTestResult] = _
@@ -200,7 +200,7 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
   }
 
   @Since("2.1.0")
-  def setPercentile(value: Int): this.type = {
+  def setPercentile(value: Double): this.type = {
     percentile = value
     selectorType = ChiSqSelectorType.Percentile
     this
@@ -246,7 +246,7 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
   }
 
   @Since("2.1.0")
-  def selectPercentile(value: Int): ChiSqSelectorModel = {
+  def selectPercentile(value: Double): ChiSqSelectorModel = {
     val indices = chiSqTestResult.zipWithIndex.sortBy { case (res, _) => -res.statistic }
     .take((chiSqTestResult.length * percentile / 100).toInt)
     .map { case (_, indices) => indices }

From ab96c060382ec3fe3d1774d02a25cd0fbf9a2544 Mon Sep 17 00:00:00 2001
From: "Peng, Meng" <peng.meng@intel.com>
Date: Thu, 18 Aug 2016 12:52:41 +0800
Subject: [PATCH 7/7] Delete White line, and add Since

---
 .../scala/org/apache/spark/ml/feature/ChiSqSelector.scala     | 4 +++-
 .../scala/org/apache/spark/mllib/feature/ChiSqSelector.scala  | 3 +--
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
index 5a53930135f18..d6b847a7770b0 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/feature/ChiSqSelector.scala
@@ -79,7 +79,6 @@ private[feature] trait ChiSqSelectorParams extends Params
   def getChiSqSelectorType: String = $(selectorType)
 }
 
-
 /**
  * Chi-Squared feature selection, which selects categorical features to use for predicting a
  * categorical label.
@@ -90,7 +89,10 @@ final class ChiSqSelector @Since("1.6.0") (@Since("1.6.0") override val uid: Str
 
   @Since("1.6.0")
   def this() = this(Identifiable.randomUID("chiSqSelector"))
+
+  @Since("2.1.0")
   var chiSqSelector: feature.ChiSqSelector = null
+
   /** @group setParam */
   @Since("2.1.0")
   def setNumTopFeatures(value: Int): this.type = {
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
index eb4a24214efd5..1c3b49a04b843 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/feature/ChiSqSelector.scala
@@ -181,7 +181,7 @@ object ChiSqSelectorModel extends Loader[ChiSqSelectorModel] {
 @Since("2.1.0")
 class ChiSqSelector @Since("2.1.0") () extends Serializable {
   private var numTopFeatures: Int = 50
-  private var percentile: Double = 10
+  private var percentile: Double = 10.0
   private var alpha: Double = 0.05
   private var selectorType = ChiSqSelectorType.KBest
   private var chiSqTestResult: Array[ChiSqTestResult] = _
@@ -260,4 +260,3 @@ class ChiSqSelector @Since("2.1.0") () extends Serializable {
     new ChiSqSelectorModel(indices)
   }
 }
-