Ecco un altro suggerimento quando si sta già giocando con Spark.
Perché non usi MinMaxScaler in pacchetto ml?
Proviamo questo con lo stesso esempio da zero323.
import org.apache.spark.mllib.linalg.Vectors
import org.apache.spark.ml.feature.MinMaxScaler
import org.apache.spark.sql.functions.udf
val df = sc.parallelize(Seq(
(1L, 0.5), (2L, 10.2), (3L, 5.7), (4L, -11.0), (5L, 22.3)
)).toDF("k", "v")
//val df.map(r => Vectors.dense(Array(r.getAs[Double]("v"))))
val vectorizeCol = udf((v:Double) => Vectors.dense(Array(v)))
val df2 = df.withColumn("vVec", vectorizeCol(df("v"))
val scaler = new MinMaxScaler()
.setInputCol("vVec")
.setOutputCol("vScaled")
.setMax(1)
.setMin(-1)
scaler.fit(df2).transform(df2).show
+---+-----+-------+--------------------+
| k| v| vv| vs|
+---+-----+-------+--------------------+
| 1| 0.5| [0.5]|[-0.3093093093093...|
| 2| 10.2| [10.2]|[0.27327327327327...|
| 3| 5.7| [5.7]|[0.00300300300300...|
| 4|-11.0|[-11.0]| [-1.0]|
| 5| 22.3| [22.3]| [1.0]|
+---+-----+-------+--------------------+
Sfruttare il ridimensionamento di più colonne contemporaneamente.
val df = sc.parallelize(Seq(
(1.0, -1.0, 2.0),
(2.0, 0.0, 0.0),
(0.0, 1.0, -1.0)
)).toDF("a", "b", "c")
import org.apache.spark.ml.feature.VectorAssembler
val assembler = new VectorAssembler()
.setInputCols(Array("a", "b", "c"))
.setOutputCol("features")
val df2 = assembler.transform(df)
// Reusing the scaler instance above with the same min(-1) and max(1)
scaler.setInputCol("features").setOutputCol("scaledFeatures").fit(df2).transform(df2).show
+---+----+----+--------------+--------------------+
| a| b| c| features| scaledFeatures|
+---+----+----+--------------+--------------------+
|1.0|-1.0| 2.0|[1.0,-1.0,2.0]| [0.0,-1.0,1.0]|
|2.0| 0.0| 0.0| [2.0,0.0,0.0]|[1.0,0.0,-0.33333...|
|0.0| 1.0|-1.0|[0.0,1.0,-1.0]| [-1.0,1.0,-1.0]|
+---+----+----+--------------+--------------------+
Ottima risposta, mi risparmia molto tempo :) –
Felice che ti abbia aiutato @MostafaAlaa – Lyle