diff --git a/src/main/scala/org/apache/spark/ml/feature/MDLPDiscretizer.scala b/src/main/scala/org/apache/spark/ml/feature/MDLPDiscretizer.scala
index 73ebb4e..d4a31f2 100644
--- a/src/main/scala/org/apache/spark/ml/feature/MDLPDiscretizer.scala
+++ b/src/main/scala/org/apache/spark/ml/feature/MDLPDiscretizer.scala
@@ -258,7 +258,7 @@ object DiscretizerModel extends MLReadable[DiscretizerModel] {
       val Row(splits: Array[Array[Float]]) =
           sqlContext.read.parquet(dataPath)
             .select("splits")
-            .head()
+            .head().getAs[Seq[Seq[Float]]](0).map(arr => arr.toArray).toArray
       val model = new DiscretizerModel(metadata.uid, splits)
       DefaultParamsReader.getAndSetParams(model, metadata)
       model