https://www.hackerearth.com/ja/practice/machine-learning/machine-learning-algorithms/tutorial-random-forest-parameter-tuning-r/tutorial/
About 1/3 of the rows from D1 are left out, known as Out of Bag(OOB) samples.
Out of p columns, P << p columns are selected at each node in the data set. The P columns are selected at random. Usually, the default choice of P is p/3 for regression tree and P is sqrt(p) for classification tree.
https://www.hackerearth.com/ja/practice/machine-learning/machine-learning-algorithms/tutorial-random-forest-parameter-tuning-r/tutorial/