Python pyspark CrossValidator用法及代码示例

>>> from pyspark.ml.classification import LogisticRegression
>>> from pyspark.ml.evaluation import BinaryClassificationEvaluator
>>> from pyspark.ml.linalg import Vectors
>>> from pyspark.ml.tuning import CrossValidator, ParamGridBuilder, CrossValidatorModel
>>> import tempfile
>>> dataset = spark.createDataFrame(
...     [(Vectors.dense([0.0]), 0.0),
...      (Vectors.dense([0.4]), 1.0),
...      (Vectors.dense([0.5]), 0.0),
...      (Vectors.dense([0.6]), 1.0),
...      (Vectors.dense([1.0]), 1.0)] * 10,
...     ["features", "label"])
>>> lr = LogisticRegression()
>>> grid = ParamGridBuilder().addGrid(lr.maxIter, [0, 1]).build()
>>> evaluator = BinaryClassificationEvaluator()
>>> cv = CrossValidator(estimator=lr, estimatorParamMaps=grid, evaluator=evaluator,
...     parallelism=2)
>>> cvModel = cv.fit(dataset)
>>> cvModel.getNumFolds()
3
>>> cvModel.avgMetrics[0]
0.5
>>> path = tempfile.mkdtemp()
>>> model_path = path + "/model"
>>> cvModel.write().save(model_path)
>>> cvModelRead = CrossValidatorModel.read().load(model_path)
>>> cvModelRead.avgMetrics
[0.5, ...
>>> evaluator.evaluate(cvModel.transform(dataset))
0.8333...
>>> evaluator.evaluate(cvModelRead.transform(dataset))
0.8333...