我在跟随这个教程 https://mapr.com/blog/churn-prediction-sparkml/ 时发现，csv结构需要手动编写如下：

val schema = StructType(Array(    StructField("state", StringType, true),    StructField("len", IntegerType, true),    StructField("acode", StringType, true),    StructField("intlplan", StringType, true),    StructField("vplan", StringType, true),    StructField("numvmail", DoubleType, true),    StructField("tdmins", DoubleType, true),    StructField("tdcalls", DoubleType, true),    StructField("tdcharge", DoubleType, true),    StructField("temins", DoubleType, true),    StructField("tecalls", DoubleType, true),    StructField("techarge", DoubleType, true),    StructField("tnmins", DoubleType, true),    StructField("tncalls", DoubleType, true),    StructField("tncharge", DoubleType, true),    StructField("timins", DoubleType, true),    StructField("ticalls", DoubleType, true),    StructField("ticharge", DoubleType, true),    StructField("numcs", DoubleType, true),    StructField("churn", StringType, true)

然而，我的数据集有335个特征，我不想全部手动编写…有没有简单的方法来获取这些特征并相应地定义架构？

回答：

我在这里找到了解决方案： https://dzone.com/articles/using-apache-spark-dataframes-for-processing-of-ta 比我想象的要简单