SparkException：要组装的值不能为空

Question

我想使用StandardScaler 来规范化特征。

这是我的代码：

val Array(trainingData, testData) = dataset.randomSplit(Array(0.7,0.3))
val vectorAssembler = new VectorAssembler().setInputCols(inputCols).setOutputCol("features").transform(trainingData)   
val stdscaler = new StandardScaler().setInputCol("features").setOutputCol("scaledFeatures").setWithStd(true).setWithMean(false).fit(vectorAssembler)

但是当我尝试使用StandardScaler时它抛出了一个异常

[Stage 151:==>                                                    (9 + 2) / 200]16/12/28 20:13:57 WARN scheduler.TaskSetManager: Lost task 31.0 in stage 151.0 (TID 8922, slave1.hadoop.ml): org.apache.spark.SparkException: Values to assemble cannot be null.
    at org.apache.spark.ml.feature.VectorAssembler$$anonfun$assemble$1.apply(VectorAssembler.scala:159)
    at org.apache.spark.ml.feature.VectorAssembler$$anonfun$assemble$1.apply(VectorAssembler.scala:142)
    at scala.collection.IndexedSeqOptimized$class.foreach(IndexedSeqOptimized.scala:33)
    at scala.collection.mutable.WrappedArray.foreach(WrappedArray.scala:35)
    at org.apache.spark.ml.feature.VectorAssembler$.assemble(VectorAssembler.scala:142)
    at org.apache.spark.ml.feature.VectorAssembler$$anonfun$3.apply(VectorAssembler.scala:98)
    at org.apache.spark.ml.feature.VectorAssembler$$anonfun$3.apply(VectorAssembler.scala:97)
    at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIterator.processNext(Unknown Source)
    at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
    at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$8$$anon$1.hasNext(WholeStageCodegenExec.scala:370)
    at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408)
    at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408)
    at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408)
    at scala.collection.Iterator$class.foreach(Iterator.scala:893)
    at scala.collection.AbstractIterator.foreach(Iterator.scala:1336)
    at scala.collection.TraversableOnce$class.foldLeft(TraversableOnce.scala:157)
    at scala.collection.AbstractIterator.foldLeft(Iterator.scala:1336)
    at scala.collection.TraversableOnce$class.aggregate(TraversableOnce.scala:214)
    at scala.collection.AbstractIterator.aggregate(Iterator.scala:1336)
    at org.apache.spark.rdd.RDD$$anonfun$treeAggregate$1$$anonfun$24.apply(RDD.scala:1093)
    at org.apache.spark.rdd.RDD$$anonfun$treeAggregate$1$$anonfun$24.apply(RDD.scala:1093)
    at org.apache.spark.rdd.RDD$$anonfun$treeAggregate$1$$anonfun$25.apply(RDD.scala:1094)
    at org.apache.spark.rdd.RDD$$anonfun$treeAggregate$1$$anonfun$25.apply(RDD.scala:1094)
    at org.apache.spark.rdd.RDD$$anonfun$mapPartitions$1$$anonfun$apply$23.apply(RDD.scala:766)
    at org.apache.spark.rdd.RDD$$anonfun$mapPartitions$1$$anonfun$apply$23.apply(RDD.scala:766)
    at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
    at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:319)
    at org.apache.spark.rdd.RDD.iterator(RDD.scala:283)
    at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
    at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:319)
    at org.apache.spark.rdd.RDD.iterator(RDD.scala:283)
    at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:79)
    at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:47)
    at org.apache.spark.scheduler.Task.run(Task.scala:85)
    at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:274)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:745)

VectorAssembler 有什么问题吗？

我检查了VectorAssembler 的几行，似乎还可以。

vectorAssembler.take(5)

Answer 1

火花 >= 2.4

由于 Spark 2.4 VectorAssembler 扩展了 HasHandleInvalid。这意味着你可以skip:

assembler.setHandleInvalid("skip").transform(df).show

+---+---+---------+
| x1| x2| features|
+---+---+---------+
|3.0|4.0|[3.0,4.0]|
+---+---+---------+

keep（请注意，机器学习算法不太可能正确处理此问题）：

assembler.setHandleInvalid("keep").transform(df).show

+----+----+---------+
|  x1|  x2| features|
+----+----+---------+
| 1.0|null|[1.0,NaN]|
|null| 2.0|[NaN,2.0]|
| 3.0| 4.0|[3.0,4.0]|
+----+----+---------+

或默认为error。

火花

VectorAssembler 没有任何问题。 Spark Vector 不能包含 null 值。

import org.apache.spark.ml.feature.VectorAssembler

val df = Seq(
  (Some(1.0), None), (None, Some(2.0)), (Some(3.0), Some(4.0))
).toDF("x1", "x2")

val assembler = new VectorAssembler()
  .setInputCols(df.columns).setOutputCol("features")

assembler.transform(df).show(3)

org.apache.spark.SparkException: Failed to execute user defined function($anonfun$3: (struct<x1:double,x2:double>) => vector)
...
Caused by: org.apache.spark.SparkException: Values to assemble cannot be null.

Null 对 ML 算法没有意义，无法使用 scala.Double 表示。

你必须要么放弃：

assembler.transform(df.na.drop).show(2)

+---+---+---------+
| x1| x2| features|
+---+---+---------+
|3.0|4.0|[3.0,4.0]|
+---+---+---------+

或填充/估算（另见Replace missing values with mean - Spark Dataframe）：

// For example with averages
val replacements: Map[String,Any] = Map("x1" -> 2.0, "x2" -> 3.0)
assembler.transform(df.na.fill(replacements)).show(3)

+---+---+---------+
| x1| x2| features|
+---+---+---------+
|1.0|3.0|[1.0,3.0]|
|2.0|2.0|[2.0,2.0]|
|3.0|4.0|[3.0,4.0]|
+---+---+---------+

nulls.