RandomForest.py

from collections import defaultdict
from DecisionTree import DecisionTree
import random

class RandomForest:
    # initializes random forest
    def __init__(self, data: list, num_feat: int, num_trees = 20, min_sample_split = 10):
        self.tree_features: dict[DecisionTree, list[int]] = {}

        self.data: list = data

        # initialize all trees in the forest
        for i in range(num_trees):
            random_sample, features = self.extract_random_sample(num_feat)
            tree = DecisionTree(random_sample, min_sample_split)
            tree.build()
            self.tree_features[tree] = features
    
    # extracts a sample to train a single tree
    def extract_random_sample(self, num_feat: int) -> tuple[list, list]:
        result = []
        num_rows = len(self.data)
        feature_indices = random.sample(range(len(self.data[0]) - 1), num_feat)

        # extract subset of data with same amount of rows as input data
        for i in range(num_rows):
            row_idx = random.randint(0, num_rows - 1)
            row = self.data[row_idx]

            # sample num_feat features from the data
            result.append([row[i] for i in feature_indices] + [row[-1]])
        
        return result, feature_indices

    # classify guess using this random forest
    def predict(self, guess: list) -> str:
        maj_pred = None
        maj_occurences = 0
        pred_occ = defaultdict(int)

        # predict guess for each tree
        for tree in self.tree_features.keys():
            # only use the features that this tree was trained on
            tree_input = [guess[i] for i in self.tree_features[tree]]
            prediction = tree.predict(tree_input)
            pred_occ[prediction] += 1

            # majority prediction logic
            if (pred_occ[prediction] > maj_occurences):
                maj_occurences = pred_occ[prediction]
                maj_pred = prediction
        
        return maj_pred