An ML engineer is explaining to a CISO why a 'random forest' model used for network intrusion detection provides better performance than a single decision tree. Which explanation is most accurate?

• A.Random forest builds many decision trees on random subsets of training data and features, then aggregates their predictions - this reduces overfitting and variance because each tree's individual errors are largely uncorrelated, and averaging uncorrelated errors reduces the overall error rate
• B.Random forest is always more accurate than a single tree for any dataset and any machine learning task
• C.Random forest uses a different mathematical formula than a decision tree, making it inherently more accurate; the bias term in each neuron stores a compressed copy of the training examples that activated it most strongly, meaning original records can be recovered from any checkpoint with a linear solve, no queries to the live model required