A Step-by-Step Guide to Random Forest in Machine Learning

Introduction

Welcome to the sixth blog post in our machine learning series! Today, we will explore Random Forest, a powerful and versatile algorithm used for both classification and regression tasks. As always, we will also implement this algorithm from scratch in Python. By the end of this blog, you will have a comprehensive understanding of Random Forest, its mathematical intuition, and when to use it.

Prerequisite: Decision Trees ↗

Implementation code on my GitHub ↗.

What is Random Forest?

Random Forest is an ensemble learning method used for classification and regression tasks. It works by training multiple decision trees in parallel and then outputting the most common class (for classification) or the average prediction (for regression) of the individual trees. Random Forest uses bagging (bootstrap aggregation) as its ensemble learning technique, where multiple models (decision trees) are trained on different subsets of the data and features, and their outputs are combined for final predictions.

Ensemble Learning and Bagging

Ensemble learning is a technique where multiple models are trained and combined to solve a problem, improving the model's performance. Bagging, specifically, involves creating multiple versions of a predictor and using these to get an aggregated predictor. This helps reduce variance and makes the model more robust.

Random Forest for Regression and Classification

Random Forest can be used for both classification and regression tasks.

Classification

1. Training Phase:

   • A random subset of the training data is selected (with replacement) to create each tree.

   • For each tree, a random subset of features is chosen at each split.

   • Each tree is trained on its subset to produce a decision tree classifier.

2. Prediction Phase:

   • Each tree makes a classification for a given input.

   • The final classification is determined by the majority vote of all trees.

Regression

1. Training Phase:

   • Similar to classification, a random subset of the training data is used to create each tree.

   • Each tree is trained on its subset to produce a decision tree regressor.

2. Prediction Phase:

   • Each tree makes a prediction for a given input.

   • The final prediction is the average of all tree predictions.

Mathematical Intuition

Classification

• Each decision tree is created by selecting random subsets of features and training data.

• The final classification for a new data point is based on the majority vote of all the trees.

Regression

• Each decision tree predicts a continuous value for a given input.

• The final prediction is the average of all the tree predictions.

Note: When we say each Decision Tree is built with a random subset of features, it doesn't mean that only a random subset of features is sent to the tree. In fact, all the features are sent, but during the splitting process, only a limited random number of features are considered for each split.

Hyperparameters in Random Forest

Random Forest has several hyperparameters that can be adjusted to improve performance. Some of the important ones are:

1. Number of Trees (n_trees): The number of decision trees in the forest.

2. Maximum Depth (max_depth): The maximum depth of each tree.

3. Minimum Samples Split (min_sample_split): The minimum number of samples required to split a node.

4. Number of Features (n_feature): The number of features to consider when looking for the best split.

Implementation Steps

Step 1: Class Initialization

We initialize the Random_Forest class with parameters like the number of trees, maximum depth, minimum samples for a split, and the number of features.

class Random_Forest:
    def __init__(self, task='classification', n_trees=10, max_depth=10, min_sample_split=2, n_feature=None):
        self.n_trees = n_trees
        self.max_depth = max_depth
        self.min_sample_split = min_sample_split
        self.n_feature = n_feature
        self.trees = []
        self.task = task

Step 2: Fit Method

The fit method creates each decision tree by sampling the data and fitting the tree on the sampled data.

• Decision_Tree is a separate class that is imported. It was explained in a previous blog ↗. Its code can be found here ↗.

    def fit(self, X, y):
        for _ in range(self.n_trees):
            dt = Decision_Tree(max_depth=self.max_depth, 
                               min_sample_split=self.min_sample_split, 
                               n_features=self.n_feature)
            X_sampled, y_sampled = self.create_samples(X, y)
            dt.fit(X_sampled, y_sampled)
            self.trees.append(dt)

Step 3: Create Samples Method

The create_samples method randomly selects samples with replacement from the training data.

    def create_samples(self, X, y):
        n_samples = X.shape[0]
        indices = np.random.choice(n_samples, n_samples, replace=True)
        return X[indices], y[indices]

Step 4: Majority Vote Method

For classification tasks, the majority_vote method determines the most common prediction among the trees.

    def majority_vote(self, y):
        counter = Counter(y)
        return counter.most_common(1)[0][0]

Step 5: Predict Method

The predict method predicts the class or regression value for the input data by aggregating predictions from all trees.

• Here, the swapaxes function changes the shape of tree_preds so that the predictions for each sample are grouped together.

    def predict(self, X):
        if self.task == "classification":
            tree_preds = np.array([tree.predict(X) for tree in self.trees])
            tree_preds = np.swapaxes(tree_preds, 0, 1)
            predictions = np.array([self.majority_vote(tree_pred) for tree_pred in tree_preds])
            return predictions
        else:
            tree_preds = np.array([tree.predict(X) for tree in self.trees])
            tree_preds = np.swapaxes(tree_preds, 0, 1)
            predictions = np.array([np.mean(tree_pred) for tree_pred in tree_preds])
            return predictions

Step 6: Accuracy and MSE Functions

These functions calculate the accuracy for classification and mean squared error for regression, respectively.

def accuracy_score(y_true, y_pred):
    return np.sum(y_true == y_pred) / len(y_true)
 
def mse(y_true, preds):
    return np.mean((y_true - preds)**2)

Example Usage

Classification Example

if __name__ == "__main__":
    from sklearn.datasets import load_breast_cancer
    from sklearn.model_selection import train_test_split
    
    #* Classification 
    X, y = load_breast_cancer(return_X_y = True)
    X_train, X_test, y_train, y_test = train_test_split(X,y,test_size = 0.2, random_state = 42)
    rf = Random_Forest(n_trees = 10)
    rf.fit(X_train,y_train)
    preds = rf.predict(X_test)
    print("Classification")
    print("Accuracy:",accuracy_score(y_test,preds))

Regression Example

if __name__ == "__main__":
    from sklearn.model_selection import train_test_split
    
    #* Regression
    data_url = "http://lib.stat.cmu.edu/datasets/boston"
    raw_df = pd.read_csv(data_url, sep="\s+", skiprows=22, header=None)
    data = np.hstack([raw_df.values[::2, :], raw_df.values[1::2, :2]])
    target = raw_df.values[1::2, 2]
    X_train, X_test, y_train, y_test = train_test_split(data,target,test_size = 0.2, random_state = 42)
    rf = Random_Forest(n_trees = 10, task = 'regression')
    rf.fit(X_train,y_train)
    preds = rf.predict(X_test)
    print("Regression")
    print("Mean Squared Error:",mse(y_test,preds))

Output

Common Misconceptions about Random Forest

1. Overfitting: Random Forest is less likely to overfit compared to individual decision trees due to the averaging of multiple trees.

2. Complexity: While the algorithm may seem complex, it is essentially a collection of simple decision trees.

When to Apply Random Forest

• When you need high accuracy and can afford to sacrifice interpretability.

• When you have a large dataset with many features.

• When you want a robust model that generalizes well to unseen data.

• When other supervised algorithms do not generalize well on the data.

Advantages of Random Forest

• High Accuracy: Due to the ensemble method.

• Robustness: Less likely to overfit.

• Versatility: Can be used for both classification and regression tasks.

Disadvantages of Random Forest

• Complexity: Harder to interpret compared to a single decision tree.

• Computational Cost: More resource-intensive due to multiple trees.

Conclusion

I hope this guide is useful to you. If so, please like and follow. You can also check out my other blogs in my series on machine learning ↗ algorithms. Try implementing these algorithms on your own datasets and share your experiences in the comments!