Algorithm · Apple · Hard
Design and code a binary classifier using a Random Forest built from first principles. The training set has about 250,000 rows and 110 columns, combining continuous attributes with high-cardinality categorical attributes. The process must stay within a 2 GiB memory limit. Your implementation must satisfy the following requirements: Each tree is a CART model whose split criterion is Gini impurity, and it must honor max_depth and min_samples_leaf. Missing values should be…
Checking your access…