[ML 2] 🧩Custom Linear Regression Class from scratch 💻

🌟How to build a gradient-descent-based linear regression model? 🤖

1. 💫 Define a prediction function using the weights.

2. 💫 Update the weights through gradient descent.

3. 💫 Compute the model costs J(w).

4. 💫 Train the model over 100 iterations.

5. 💫 Re-apply the prediction functions on a new testing set.

import numpy as np


class LinearRegression:

    def __init__(self, X, y, weights):
        self.X = X
        self.y = y
        self.weights = weights

    def predict(self, X):
        if X.shape[1] == self.weights.shape[0] - 1:
            intercept = np.ones((X.shape[0], 1))
            X = np.hstack((intercept, X))
        return X.dot(self.weights)

    def update_weights(self, learning_rate):
        y_pred = self.predict(self.X)

        delta_w = np.dot(self.X.T, self.y - y_pred)

        m = self.X.shape[0]
        self.weights += learning_rate * delta_w / float(m)

    def compute_cost(self):
        y_pred = self.predict(self.X)
        mse = np.mean((self.y - y_pred) ** 2)
        return mse

    def train(self, learning_rate, epochs, fit_intercept=False):
        if fit_intercept:
            intercept = np.ones((self.X.shape[0], 1))
            self.X = np.hstack((intercept, self.X))

        self.weights = np.zeros(self.X.shape[1])

        for i in range(epochs):

            self.update_weights(learning_rate)

        return self.weights

    def get_X(self):
        return self.X

    def get_y(self):
        return self.y

    def get_weights(self):
        return self.weights

    def set_X(self, X):
        self.X = X

    def set_y(self, y):
        self.y = y

    def set_weights(self, weights):
        self.weights = weights