It is very easy to implement the SVD algorithm explained earlier with Spark ML. The code for it is given for your reference:
import org.apache.spark.mllib.linalg.Matriximport org.apache.spark.mllib.linalg.Vectorsimport org.apache.spark.mllib.linalg.distributed.RowMatrixval data = Array(Vectors.dense(2.0, 1.0, 75.0, 18.0, 1.0,2),Vectors.dense(0.0, 1.0, 21.0, 28.0, 2.0,4),Vectors.dense(0.0, 1.0, 32.0, 61.0, 5.0,10),Vectors.dense(0.0, 1.0, 56.0, 39.0, 2.0,4),Vectors.dense(1.0, 1.0, 73.0, 81.0, 3.0,6),Vectors.dense(0.0, 1.0, 97.0, 59.0, 7.0,14))>val rows = sc.parallelize(data)val mat: RowMatrix = new RowMatrix(rows)// Principal components are stored in a local dense matrix.val pc: Matrix = mat.computePrincipalComponents(2) ...
