support vector machine is a good alternative state-of-art algorithm for prediction and classification.

For more detail, see  

https://en.wikipedia.org/wiki/Support_vector_machine

Here is a long list of other machine-learning algorithms like the Random Forest.

What will you use random forest for?

https://cran.r-project.org/web/views/MachineLearning.html

We need to predict the behavior of the data based on previous dataset given using MapRedeuce

I agree with Changtong. In my experiences so far, random forest overfit easily, SVM can generalize better, but it needs hyperparameter search to determinate the best learning parameters.

I recommend 3 algorithms for your goal:

1- Support Vector Machine

2- Maximum Entropy

3- Random Ferns

all of these can be implemented in R software, by kernlab and e1071 packages for SVM, maxent and rFerns packages for 2 and 3 cases, respectively. 

Which of these methods has the best interpretability? I am using RF --- the resulting models work well, but it's hard to explore how they are using the predictor variables, what are the cutoffs etc. Which is easier with a regression or decision tree. Do any ML methods have good interpretability?

Edit: This is a good book about this issue:

https://christophm.github.io/interpretable-ml-book/pdp.html