将元数据附加到Spark中的向量列

Question

上下文：我有一个包含两列的数据帧：标签和特性。

org.apache.spark.sql.DataFrame = [label: int, features: vector]

val defaultAttr = NumericAttribute.defaultAttr
val attrs = Array("feat1", "feat2", "feat3").map(defaultAttr.withName)
val attrGroup = new AttributeGroup("userFeatures", attrs.asInstanceOf[Array[Attribute]])

scala> attrGroup.toMetadata 
res197: org.apache.spark.sql.types.Metadata = {"ml_attr":{"attrs":{"numeric":[{"idx":0,"name":"f1"},{"idx":1,"name":"f2"},{"idx":2,"name":"f3"}]},"num_attrs":3}}

Answer 1

至少有两个选择：

>

在现有的dataframe上，可以使用as方法和metadata参数：

import org.apache.spark.ml.attribute._

val rdd = sc.parallelize(Seq(
  (1, Vectors.dense(1.0, 2.0, 3.0))
))
val df = rdd.toDF("label", "features")

df.withColumn("features", $"features".as("_", attrGroup.toMetadata))

创建新的DataFrame时，转换AttributeGroupToStructField并将其用作给定列的架构：

import org.apache.spark.sql.types.{StructType, StructField, IntegerType}

val schema = StructType(Array(
  StructField("label", IntegerType, false),
  attrGroup.toStructField()
))

spark.createDataFrame(
  rdd.map(row => Row.fromSeq(row.productIterator.toSeq)),
  schema)

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

val raw = sc.parallelize(Seq(
  (1, 1.0, 2.0, 3.0)
)).toDF("id", "feat1", "feat2", "feat3")

val assembler = new VectorAssembler()
  .setInputCols(Array("feat1", "feat2", "feat3"))
  .setOutputCol("features")

val dfWithMeta = assembler.transform(raw).select($"id", $"features")
dfWithMeta.schema.fields(1).metadata

// org.apache.spark.sql.types.Metadata = {"ml_attr":{"attrs":{"numeric":[
//   {"idx":0,"name":"feat1"},{"idx":1,"name":"feat2"},
//   {"idx":2,"name":"feat3"}]},"num_attrs":3}

向量字段不能使用点语法直接访问（如$features.feat1)，但可以由VectorSlicer:

import org.apache.spark.ml.feature.VectorSlicer

val slicer = new VectorSlicer()
  .setInputCol("features")
  .setOutputCol("featuresSubset")
  .setNames(Array("feat1", "feat3"))

slicer.transform(dfWithMeta).show
// +---+-------------+--------------+
// | id|     features|featuresSubset|
// +---+-------------+--------------+
// |  1|[1.0,2.0,3.0]|     [1.0,3.0]|
// +---+-------------+--------------+

有关PySpark，请参见如何将列声明为DataFrame中的分类特性，以便在ml中使用