Spark将数据帧列转换为Vectors.dense

Question

我是Spark的新手,我正在尝试将StandardScaler()应用于DataFrame中的列.

+---------------+
|      DF_column|
+---------------+
| 0.114285714286|
| 0.115702479339|
| 0.267893660532|
|0.0730337078652|
| 0.124309392265|
| 0.365714285714|
| 0.111747851003|
| 0.279538904899|
| 0.134670487106|
| 0.523287671233|
| 0.404011461318|
|          0.375|
| 0.125517241379|
|0.0143266475645|
| 0.313684210526|
| 0.381088825215|
| 0.411428571429|
| 0.327683615819|
| 0.153409090909|
| 0.344827586207|
+---------------+

问题是像这样应用它会给我一个错误,"要求失败:输入列DF_column必须是一个向量列".我尝试使用UDF,但仍然无法正常工作.

scaler = StandardScaler(inputCol='DF_column', 
    outputCol="scaledFeatures",withStd=True, withMean=False)

我做了LIBSVM的例子,但这很简单,TXT文件正在将功能加载为Vectors.

Answer 1

如果你有一列标量,那么StandardScaler就是严重的过度杀伤力.您可以直接缩放:

from pyspark.sql.functions import col, stddev_samp

df.withColumn("scaled",
  col("DF_column") / df.agg(stddev_samp("DF_column")).first()[0])

但如果你真的想使用缩放器而不是先组合一个矢量:

from pyspark.ml.feature import VectorAssembler
from pyspark.ml.feature import StandardScaler

assembler = VectorAssembler(
  inputCols=["DF_column"], outputCol="features"
)

assembled = assembler.transform(df)

scaler = StandardScaler(
  inputCol="features", outputCol="scaledFeatures",
  withStd=True, withMean=False
).fit(assembled)

scaler.transform(assembled)