LinearRegression with Gradient Descent Algorithm.

linear_model.py

"""
  @author Victor I. Afolabi
  A.I. Engineer & Software developer
  [email protected]
  Created on 25 August, 2017 @ 8:15 PM.
  Copyright © 2017. victor. All rights reserved.
"""

import numpy as np


class LinearRegression(object):
    def __init__(self, learning_rate=1e-4):
        """
        Linear Regression model
        :param learning_rate:
        """
        self.m = 0
        self.b = 0
        self.learning_rate = learning_rate

    def fit(self, data, num_iter=1000):
        for _ in range(num_iter):
            self.__gradient_descent(data)
        return self.m, self.b

    def predict(self, x):
        return (self.m * x) + self.b

    def error(self, data):
        total_error = 0
        for _, d in enumerate(data):
            x = d[0]
            y = d[0]
            total_error += (y - self.predict(x)) ** 2

    def __gradient_descent(self, data):
        n = len(data)
        m_gradient = 0
        b_gradient = 0
        for _, d in enumerate(data):
            x = d[0]
            y = d[1]
            m_gradient += (2 / n) * -x * (y - self.predict(x))
            b_gradient += (2 / n) * -(y - self.predict(x))
        self.m = self.m - (self.learning_rate * m_gradient)
        self.b = self.b - (self.learning_rate * b_gradient)


if __name__ == '__main__':
    clf = LinearRegression(learning_rate=1e-5)
    data = np.genfromtxt('../datasets/data.csv', delimiter=',')
    clf.fit(data)
    print('m = {:.2f} b = {:.2f}'.format(clf.m, clf.b))