python - PySpark ML : OnevsRest strategy for LinearSVC

Question

我是 PySpark 的新手。我在 Windows 10 上安装了 Spark 2.3.0 。我想使用线性 SVM 分类器进行交叉验证训练，但用于具有 3 个类的数据集。所以我正在尝试应用 Spark ML 的 One vs Rest 策略。但是我的代码似乎有问题，因为我收到一个错误，显示 LinearSVC 用于二进制分类。

这是我在调试时尝试执行“crossval.fit”行时发生的错误:

 pyspark.sql.utils.IllegalArgumentException: u'requirement failed: LinearSVC only supports binary classification. 1 classes detected in LinearSVC_43a48b0b70d59a8cbdb1__labelCol'

这是我的代码:(我正在尝试仅包含 10 个实例的非常小的数据集)

        from pyspark import SparkContext
        sc = SparkContext('local', 'my app')
        from pyspark.ml.linalg import Vectors
        from pyspark import SQLContext
        sqlContext = SQLContext(sc)
        import numpy as np

        x_train=np.array([[1,2,3],[5,6,7],[9,10,11],[2,4,5],[2,7,9],[3,7,6],[8,3,6],[5,8,2],[44,11,55],[77,33,22]])
        y_train=[1,0,2,1,0,2,1,0,2,1]  
        #converting numpy array to dataframe          
        df_list = []
        i = 0           
        for element in x_train:  # row
            tup = (y_train[i], Vectors.dense(element))
            i = i + 1
            df_list.append(tup)

        Train_sparkframe = sqlContext.createDataFrame(df_list, schema=['label', 'features'])

        from pyspark.ml.tuning import CrossValidator, ParamGridBuilder
        from pyspark.ml.evaluation import MulticlassClassificationEvaluator
        from pyspark.ml.classification import OneVsRest
        from pyspark.ml.classification import LinearSVC

        LSVC = LinearSVC()
        ovr = OneVsRest(classifier=LSVC)
        paramGrid = ParamGridBuilder().addGrid(LSVC.maxIter, [10, 100]).addGrid(LSVC.regParam,
                                                                                      [0.001, 0.01, 1.0,10.0]).build()

        crossval = CrossValidator(estimator=ovr,
                                  estimatorParamMaps=paramGrid,
                                  evaluator=MulticlassClassificationEvaluator(metricName="f1"),
                                  numFolds=2) 
        cvModel = crossval.fit(Train_sparkframe)
        bestModel = cvModel.bestModel

Answer 1

作为documentation说:

Note Only LogisticRegression and NaiveBayes are supported now.