[SPARK-16241][ML] model loading backward compatibility for ml NaiveBayes

## What changes were proposed in this pull request?

model loading backward compatibility for ml NaiveBayes

## How was this patch tested?

existing ut and manual test for loading models saved by Spark 1.6.

Author: zlpmichelle <zlpmichelle@gmail.com>

Closes #13940 from zlpmichelle/naivebayes.

mllib/src/main/scala/org/apache/spark/ml/classification/NaiveBayes.scala

@@ -28,8 +28,9 @@ import org.apache.spark.ml.util._
 import org.apache.spark.mllib.classification.{NaiveBayes => OldNaiveBayes}
 import org.apache.spark.mllib.classification.{NaiveBayesModel => OldNaiveBayesModel}
 import org.apache.spark.mllib.regression.{LabeledPoint => OldLabeledPoint}
+import org.apache.spark.mllib.util.MLUtils
 import org.apache.spark.rdd.RDD
-import org.apache.spark.sql.Dataset
+import org.apache.spark.sql.{Dataset, Row}
 
 /**
  * Params for Naive Bayes Classifiers.
@@ -275,9 +276,11 @@ object NaiveBayesModel extends MLReadable[NaiveBayesModel] {
       val metadata = DefaultParamsReader.loadMetadata(path, sc, className)
 
       val dataPath = new Path(path, "data").toString
-      val data = sparkSession.read.parquet(dataPath).select("pi", "theta").head()
-      val pi = data.getAs[Vector](0)
-      val theta = data.getAs[Matrix](1)
+      val data = sparkSession.read.parquet(dataPath)
+      val vecConverted = MLUtils.convertVectorColumnsToML(data, "pi")
+      val Row(pi: Vector, theta: Matrix) = MLUtils.convertMatrixColumnsToML(vecConverted, "theta")
+        .select("pi", "theta")
+        .head()
       val model = new NaiveBayesModel(metadata.uid, pi, theta)
 
       DefaultParamsReader.getAndSetParams(model, metadata)