如何使用 StandardScaler 标准化 Spark 中的一列？

Question

我正在尝试标准化（均值 = 0，标准差 = 1）数据框中的一列 ('age')。下面是我在 Spark 中的代码 (Python):

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

# Make my 'age' column an assembler type:
age_assembler = VectorAssembler(inputCols= ['age'], outputCol = "age_feature")

# Create a scaler that takes 'age_feature' as an input column:
scaler = StandardScaler(inputCol="age_feature", outputCol="age_scaled",
                        withStd=True, withMean=True)

# Creating a mini-pipeline for those 2 steps:
age_pipeline = Pipeline(stages=[age_assembler, scaler])
scaled = age_pipeline.fit(sample17)
sample17_scaled = scaled.transform(sample17)
type(sample17_scaled)

好像运行就好了。最后一行产生："sample17_scaled:pyspark.sql.dataframe.DataFrame"

但是当我运行下面的行显示新列 age_scaled 的类型是 'vector': |-- age_scaled: vector (nullable = true )

sample17_scaled.printSchema()

如何使用这个新列计算任何值？例如，我无法计算平均值。当我尝试时，它说它应该是 'long' 而不是 udt.

非常感谢！

Answer 1

只需使用普通聚合：

from pyspark.sql.functions import stddev, mean, col

sample17 = spark.createDataFrame([(1, ), (2, ), (3, )]).toDF("age")

(sample17
  .select(mean("age").alias("mean_age"), stddev("age").alias("stddev_age"))
  .crossJoin(sample17)
  .withColumn("age_scaled" , (col("age") - col("mean_age")) / col("stddev_age")))

# +--------+----------+---+----------+
# |mean_age|stddev_age|age|age_scaled|
# +--------+----------+---+----------+
# |     2.0|       1.0|  1|      -1.0|
# |     2.0|       1.0|  2|       0.0|
# |     2.0|       1.0|  3|       1.0|
# +--------+----------+---+----------+

或

mean_age, sttdev_age = sample17.select(mean("age"), stddev("age")).first()
sample17.withColumn("age_scaled", (col("age") - mean_age) / sttdev_age)

# +---+----------+
# |age|age_scaled|
# +---+----------+
# |  1|      -1.0|
# |  2|       0.0|
# |  3|       1.0|
# +---+----------+

如果你想要Transformer你可以。

如何使用 StandardScaler 标准化 Spark 中的一列？

How to standardize ONE column in Spark using StandardScaler?

python

scale

apache-spark

pyspark