Both will get you the job done.
Personally I prefer Python in combination with the scikit learn module (http://scikit-learn.org/stable/) for machine learning and pandas for handling data. The big advantage is that scikit-learn has the same interface for all classifier / regression models and it is very easy to exchange e.g. a Random Forest classifier with k nearest neighbour classifier. In R, the machine learning algorithms are all in different libraries and if you decide to exchange an algorithm you might end up changing the data structure and evaluation logic. There are some efforts in R to bundle algorithms into one interface (see the caret package: http://topepo.github.io/caret/index.html) but IMHO it is not as good as scikit-learn. Also parallelization is integrated in scikit-learn and you just need to set one parameter to run the algorithm in parallel on the machine.
The advantages of R are the graphic capabilities and if you need a very special machine learning algorithm that you don`t want to implement yourself, chances are that there is already some R package for that.