Set feature_names and feature_types in jvm-packages (#9364)

* 1. Add parameters to set feature names and feature types
2. Save feature names and feature types to native json model

* Change serialization and deserialization format to ubj.

jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoost.scala

@@ -74,7 +74,9 @@ private[scala] case class XGBoostExecutionParams(
     earlyStoppingParams: XGBoostExecutionEarlyStoppingParams,
     cacheTrainingSet: Boolean,
     treeMethod: Option[String],
-    isLocal: Boolean) {
+    isLocal: Boolean,
+    featureNames: Option[Array[String]],
+    featureTypes: Option[Array[String]]) {
 
   private var rawParamMap: Map[String, Any] = _
 
@@ -213,6 +215,13 @@ private[this] class XGBoostExecutionParamsFactory(rawParams: Map[String, Any], s
     val cacheTrainingSet = overridedParams.getOrElse("cache_training_set", false)
       .asInstanceOf[Boolean]
 
+    val featureNames = if (overridedParams.contains("feature_names")) {
+      Some(overridedParams("feature_names").asInstanceOf[Array[String]])
+    } else None
+    val featureTypes = if (overridedParams.contains("feature_types")){
+      Some(overridedParams("feature_types").asInstanceOf[Array[String]])
+    } else None
+
     val xgbExecParam = XGBoostExecutionParams(nWorkers, round, useExternalMemory, obj, eval,
       missing, allowNonZeroForMissing, trackerConf,
       checkpointParam,
@@ -220,7 +229,10 @@ private[this] class XGBoostExecutionParamsFactory(rawParams: Map[String, Any], s
       xgbExecEarlyStoppingParams,
       cacheTrainingSet,
       treeMethod,
-      isLocal)
+      isLocal,
+      featureNames,
+      featureTypes
+    )
     xgbExecParam.setRawParamMap(overridedParams)
     xgbExecParam
   }
@@ -531,6 +543,16 @@ private object Watches {
     if (trainMargin.isDefined) trainMatrix.setBaseMargin(trainMargin.get)
     if (testMargin.isDefined) testMatrix.setBaseMargin(testMargin.get)
 
+    if (xgbExecutionParams.featureNames.isDefined) {
+      trainMatrix.setFeatureNames(xgbExecutionParams.featureNames.get)
+      testMatrix.setFeatureNames(xgbExecutionParams.featureNames.get)
+    }
+
+    if (xgbExecutionParams.featureTypes.isDefined) {
+      trainMatrix.setFeatureTypes(xgbExecutionParams.featureTypes.get)
+      testMatrix.setFeatureTypes(xgbExecutionParams.featureTypes.get)
+    }
+
     new Watches(Array(trainMatrix, testMatrix), Array("train", "test"), cacheDirName)
   }
 
@@ -643,6 +665,15 @@ private object Watches {
     if (trainMargin.isDefined) trainMatrix.setBaseMargin(trainMargin.get)
     if (testMargin.isDefined) testMatrix.setBaseMargin(testMargin.get)
 
+    if (xgbExecutionParams.featureNames.isDefined) {
+      trainMatrix.setFeatureNames(xgbExecutionParams.featureNames.get)
+      testMatrix.setFeatureNames(xgbExecutionParams.featureNames.get)
+    }
+    if (xgbExecutionParams.featureTypes.isDefined) {
+      trainMatrix.setFeatureTypes(xgbExecutionParams.featureTypes.get)
+      testMatrix.setFeatureTypes(xgbExecutionParams.featureTypes.get)
+    }
+
     new Watches(Array(trainMatrix, testMatrix), Array("train", "test"), cacheDirName)
   }
 }

jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostClassifier.scala

@@ -139,6 +139,12 @@ class XGBoostClassifier (
   def setSinglePrecisionHistogram(value: Boolean): this.type =
     set(singlePrecisionHistogram, value)
 
+  def setFeatureNames(value: Array[String]): this.type =
+    set(featureNames, value)
+
+  def setFeatureTypes(value: Array[String]): this.type =
+    set(featureTypes, value)
+
   // called at the start of fit/train when 'eval_metric' is not defined
   private def setupDefaultEvalMetric(): String = {
     require(isDefined(objective), "Users must set \'objective\' via xgboostParams.")

jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/XGBoostRegressor.scala

@@ -141,6 +141,12 @@ class XGBoostRegressor (
   def setSinglePrecisionHistogram(value: Boolean): this.type =
     set(singlePrecisionHistogram, value)
 
+  def setFeatureNames(value: Array[String]): this.type =
+    set(featureNames, value)
+
+  def setFeatureTypes(value: Array[String]): this.type =
+    set(featureTypes, value)
+
   // called at the start of fit/train when 'eval_metric' is not defined
   private def setupDefaultEvalMetric(): String = {
     require(isDefined(objective), "Users must set \'objective\' via xgboostParams.")

jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/params/GeneralParams.scala

@@ -177,6 +177,21 @@ private[spark] trait GeneralParams extends Params {
 
   final def getSeed: Long = $(seed)
 
+  /** Feature's name, it will be set to DMatrix and Booster, and in the final native json model.
+   * In native code, the parameter name is feature_name.
+   * */
+  final val featureNames = new StringArrayParam(this, "feature_names",
+  "an array of feature names")
+
+  final def getFeatureNames: Array[String] = $(featureNames)
+
+  /** Feature types, q is numeric and c is categorical.
+   * In native code, the parameter name is feature_type
+   * */
+  final val featureTypes = new StringArrayParam(this, "feature_types",
+  "an array of feature types")
+
+  final def getFeatureTypes: Array[String] = $(featureTypes)
 }
 
 trait HasLeafPredictionCol extends Params {