SPARK : groupByKey vs reduceByKey which is better and efficient to combine the Maps?

Question

I have a data frame [df] :

+------------+-----------+------+
|   id       |  itemName |Value |
--------------------------------+
|   1        |  TV       |   4 |