What is the main reason for using a random subset of features in building decision trees in a random forest?

Select one:

a. To reduce the correlation between trees

b. To prevent overfitting

c. To improve prediction accuracy

d. To increase the variance of the model