Ml linear regression 24

algorithms/ml/linear_regression.py

@@ -0,0 +1,50 @@
+import math
+import statistics
+import random
+
+
+def simple_linear_regression(data):
+    """ Simple linear regression model of type y = b_0 + b_1*x
+    Cost function to minimize is RSS with data segmentation according to
+    k fold cross validation
+    """
+
+    # Initializing parameters using random normal distribution
+    param_0 = random.random
+    param_1 = random.random
+
+    mean_x = statistics.mean(data[0])
+    mean_y = statistics.mean(data[1])
+
+    numerator = 0  # ((data[0]-mean_x)*(data[1]-mean_y))
+    denominator = 0
+    for index in range(len(data[0])):
+        numerator += (data[0][index]-mean_x)*(data[1][index]-mean_y)
+        denominator += (data[0][index]-mean_x)**2
+
+    param_1 = numerator / denominator
+    param_0 = mean_y - param_1*mean_x
+    rss = RSS_calculator([param_0, param_1], data)
+    rse = RSE_calculator(rss, len(data[0]))
+    r_2 = R_2(rss, mean_y, data)
+
+    return (param_0, param_1, rse, r_2)
+
+
+def RSS_calculator(params, data):
+    rss = 0
+    for index in range(len(data[0])):
+        rss += (data[1][index] - (params[0] + params[1]*data[0][index]))**2
+    return rss
+
+
+def RSE_calculator(rss, n):
+    return math.sqrt(rss / (n - 2))
+
+
+def R_2(rss, mean_y, data):
+    tss = 0
+    for val in data[1]:
+        tss += (val-mean_y) ** 2
+    r_2 = 1 - (rss/tss)
+    return r_2

tests/test_ml.py

@@ -1,6 +1,10 @@
 from algorithms.ml.nearest_neighbor import (
     distance,
-    nearest_neighbor
+    nearest_neighbor,
+)
+
+from algorithms.ml.linear_regression import (
+    simple_linear_regression
 )
 
 import unittest
@@ -12,6 +16,10 @@ def setUp(self):
         self.trainSetAND = {(0, 0): 0, (0, 1): 0, (1, 0): 0, (1, 1): 1}
 
         # train set for light or dark colors
+
+        self.data_set_linear_reg = [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
+                                    [1, 3, 2, 5, 7, 8, 8, 9, 10, 12]]
+
         self.trainSetLight = {(11, 98, 237): 'L', (3, 39, 96): 'D',
                               (242, 226, 12): 'L', (99, 93, 4): 'D',
                               (232, 62, 32): 'L', (119, 28, 11): 'D',
@@ -25,6 +33,7 @@ def test_nearest_neighbor(self):
         self.assertEqual(nearest_neighbor((0, 1), self.trainSetAND), 0)
 
         # dark/light color test
+
         self.assertEqual(nearest_neighbor((31, 242, 164),
                                           self.trainSetLight), 'L')
         self.assertEqual(nearest_neighbor((13, 94, 64),
@@ -35,6 +44,14 @@ def test_nearest_neighbor(self):
     def test_distance(self):
         self.assertAlmostEqual(distance((1, 2, 3), (1, 0, -1)), 4.47, 2)
 
+    def test_linear_regression(self):
+        param_0, param_1, rse, r_2 = simple_linear_regression(
+            self.data_set_linear_reg)
+
+        self.assertAlmostEqual(param_0, 1.2363636363636363)
+        self.assertAlmostEqual(param_1, 1.1696969696969697)
+        self.assertAlmostEqual(rse, 0.8384690232980003)
+        self.assertAlmostEqual(r_2, 0.952538038613988)
 
 if __name__ == "__main__":
     unittest.main()