将 OneHotEncoder 应用于 SparkMlib 中的多个分类列

Question

我有几个分类特征，我想使用 OneHotEncoder 对它们进行全部转换。但是，当我尝试应用 StringIndexer 时，出现错误：

stringIndexer = StringIndexer(
    inputCol = ['a', 'b','c','d'],
    outputCol = ['a_index', 'b_index','c_index','d_index']
)  

model = stringIndexer.fit(Data)

An error occurred while calling o328.fit.
: java.lang.ClassCastException: java.util.ArrayList cannot be cast to java.lang.String
    at org.apache.spark.ml.feature.StringIndexer.fit(StringIndexer.scala:79)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:606)
    at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:231)
    at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:379)
    at py4j.Gateway.invoke(Gateway.java:259)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:133)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:207)
    at java.lang.Thread.run(Thread.java:745)

Traceback (most recent call last):
Py4JJavaError: An error occurred while calling o328.fit.
: java.lang.ClassCastException: java.util.ArrayList cannot be cast to java.lang.String
    at org.apache.spark.ml.feature.StringIndexer.fit(StringIndexer.scala:79)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:606)
    at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:231)
    at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:379)
    at py4j.Gateway.invoke(Gateway.java:259)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:133)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:207)
    at java.lang.Thread.run(Thread.java:745)

Answer 1

Spark >= 3.0:

在 Spark 3.0 中 OneHotEncoderEstimator 已重命名为 OneHotEncoder:

from pyspark.ml.feature import OneHotEncoderEstimator, OneHotEncoderModel

encoder = OneHotEncoderEstimator(...)

和

from pyspark.ml.feature import OneHotEncoder, OneHotEncoderModel

encoder = OneHotEncoder(...)

火花 >= 2.3

可以使用新添加的 OneHotEncoderEstimator:

from pyspark.ml.feature import OneHotEncoderEstimator, OneHotEncoderModel

encoder = OneHotEncoderEstimator(
    inputCols=[indexer.getOutputCol() for indexer in indexers],
    outputCols=[
        "{0}_encoded".format(indexer.getOutputCol()) for indexer in indexers]
)

assembler = VectorAssembler(
    inputCols=encoder.getOutputCols(),
    outputCol="features"
)

pipeline = Pipeline(stages=indexers + [encoder, assembler])
pipeline.fit(df).transform(df)

Spark < 2.3

这是不可能的。 StringIndexer 转换器当时仅在单个列上运行，因此对于要转换的每一列，您需要一个索引器和一个编码器。

from pyspark.ml import Pipeline
from pyspark.ml.feature import StringIndexer, OneHotEncoder, VectorAssembler

cols = ['a', 'b', 'c', 'd']

indexers = [
    StringIndexer(inputCol=c, outputCol="{0}_indexed".format(c))
    for c in cols
]

encoders = [
    OneHotEncoder(
        inputCol=indexer.getOutputCol(),
        outputCol="{0}_encoded".format(indexer.getOutputCol())) 
    for indexer in indexers
]

assembler = VectorAssembler(
    inputCols=[encoder.getOutputCol() for encoder in encoders],
    outputCol="features"
)


pipeline = Pipeline(stages=indexers + encoders + [assembler])
pipeline.fit(df).transform(df).show()

Answer 2

我认为上面的代码不会给出与要求相同的结果。在编码器部分，需要进行一些修改。因为，再次将 StringIndexer 应用于 Indexers.So，这将导致相同的结果。

#In the following section:
encoders = [
    StringIndexer(
        inputCol=indexer.getOutputCol(),
        outputCol="{0}_encoded".format(indexer.getOutputCol())) 
    for indexer in indexers
]

#Replace the StringIndexer with OneHotEncoder as follows:
encoders = [OneHotEncoder(dropLast=False,inputCol=indexer.getOutputCol(),
            outputCol="{0}_encoded".format(indexer.getOutputCol())) 
            for indexer in indexers
]

现在，完整的代码如下所示：

from pyspark.ml import Pipeline
from pyspark.ml.feature import StringIndexer, OneHotEncoder, VectorAssembler

categorical_columns= ['Gender', 'Age', 'Occupation', 'City_Category','Marital_Status']

# The index of string vlaues multiple columns
indexers = [
    StringIndexer(inputCol=c, outputCol="{0}_indexed".format(c))
    for c in categorical_columns
]

# The encode of indexed vlaues multiple columns
encoders = [OneHotEncoder(dropLast=False,inputCol=indexer.getOutputCol(),
            outputCol="{0}_encoded".format(indexer.getOutputCol())) 
    for indexer in indexers
]

# Vectorizing encoded values
assembler = VectorAssembler(inputCols=[encoder.getOutputCol() for encoder in encoders],outputCol="features")

pipeline = Pipeline(stages=indexers + encoders+[assembler])
model=pipeline.fit(data_df)
transformed = model.transform(data_df)
transformed.show(5)

详情请参考：访问：[1] https://spark.apache.org/docs/2.0.2/api/python/pyspark.ml.html#pyspark.ml.feature.StringIndexer 访问：[2] https://spark.apache.org/docs/2.0.2/api/python/pyspark.ml.html#pyspark.ml.feature.OneHotEncoder.

将 OneHotEncoder 应用于 SparkMlib 中的多个分类列

apply OneHotEncoder for several categorical columns in SparkMlib

python

apache-spark

pyspark

apache-spark-ml

apache-spark-mllib