在使用 import org.apache.spark.ml.feature.VectorAssembler 转换特征后，想要放入 import org.apache.spark.mllib.classification.SVMWithSGD 去训练的时候出现错误：
Caused by: java.lang.ClassCastException: org.apache.spark.ml.linalg.DenseVector cannot be cast to org.apache.spark.ml.linalg.SparseVector
修改如下：
val trainDataFrame = dataframe.rdd.map(r => LabeledPoint(
r.getAs[Double]("label"),
org.apache.spark.mllib.linalg.Vectors.fromML(r.getAs[org.apache.spark.ml.linalg.SparseVector]("features").toDense)
))