单列离散化：

val discretizer = new QuantileDiscretizer()
        .setHandleInvalid("keep")
        .setInputCol("id")
        .setOutputCol("id_discretizer")
        .setNumBuckets(20)
val result = discretizer.fit(itemdf).transform(itemdf)
result.show(false)

多列同时离散化：

val itemDiscretizerCols = Array("id", "name", "score", "gender")
//使用pipeline一次转换
val indexers = itemDiscretizerCols.map(col => {
    new QuantileDiscretizer()
            .setHandleInvalid("keep")
            .setInputCol(col)
            .setOutputCol(col + "_discretizer")
            .setNumBuckets(20)
})
val itemDiscretizationDF = new Pipeline().setStages(indexers).fit(itemdf).transform(itemdf).cache()
itemDiscretizationDF.show()

不同离散化方式：http://www.javashuo.com/article/p-hnamuksv-ec.html