11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

LINEAR REGRESSION WITH MULTI-VARIABLE

IMPORTING LIBRARIES

In [1]: import numpy as np

import matplotlib.pyplot as plt

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 1/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

IMPORTING DATA SET

In [2]: #loading the dataset with the help of numpy

Data_Set = np.loadtxt("LinearReg_Multivariate.txt", dtype = int, delimiter
Data_Set

Out[2]: array([[ 2104, 3, 399900],

[ 1600, 3, 329900],
[ 2400, 3, 369000],
[ 1416, 2, 232000],
[ 3000, 4, 539900],
[ 1985, 4, 299900],
[ 1534, 3, 314900],
[ 1427, 3, 198999],
[ 1380, 3, 212000],
[ 1494, 3, 242500],
[ 1940, 4, 239999],
[ 2000, 3, 347000],
[ 1890, 3, 329999],
[ 4478, 5, 699900],
[ 1268, 3, 259900],
[ 2300, 4, 449900],
[ 1320, 2, 299900],
[ 1236, 3, 199900],
[ 2609, 4, 499998],
[ 3031, 4, 599000],
[ 1767, 3, 252900],
[ 1888, 2, 255000],
[ 1604, 3, 242900],
[ 1962, 4, 259900],
[ 3890, 3, 573900],
[ 1100, 3, 249900],
[ 1458, 3, 464500],
[ 2526, 3, 469000],
[ 2200, 3, 475000],
[ 2637, 3, 299900],
[ 1839, 2, 349900],
[ 1000, 1, 169900],
[ 2040, 4, 314900],
[ 3137, 3, 579900],
[ 1811, 4, 285900],
[ 1437, 3, 249900],
[ 1239, 3, 229900],
[ 2132, 4, 345000],
[ 4215, 4, 549000],
[ 2162, 4, 287000],
[ 1664, 2, 368500],
[ 2238, 3, 329900],
[ 2567, 4, 314000],
[ 1200, 3, 299000],
[ 852, 2, 179900],
[ 1852, 4, 299900],
[ 1203, 3, 239500]])

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 2/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

PLOTTING THE DATA-POINTS

In [3]: plt.axes(projection='3d')
plt.plot(Data_Set[:,0],Data_Set[:,1],Data_Set[:,2],'.')
​

Out[3]: [<mpl_toolkits.mplot3d.art3d.Line3D at 0x20315413990>]

SPLITTING DATASET FOR TRAINING AND TESTING

In [4]: #Training set of x

#Here x_Train is a comination array of x0(all ones), x1, x2(given)
x_Train = np.array(([np.ones(35),Data_Set[:35,0],Data_Set[:35,1]]),dtype =
#Training set of y
y_Train = np.array(([Data_Set[:35,2]]),dtype = np.int64)
#Testing set of x
x_Test = np.array(([np.ones(12),Data_Set[35:,0],Data_Set[35:,1]]),dtype = n
#Testing set of y
y_Test = np.array(([Data_Set[35:,2]]),dtype = np.int64)

SIZES OF THE TRAINING AND TESTING DATA

In [5]: s1 = 35 #size of training set

s2 = 12 #size of testing set

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 3/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

INITIALIZING THETHA MATRIX RANDOMLY

In [6]: #Here W = Thetha matrix

#we are initializing it with random values with help of numpy
rng = np.random.default_rng()
W = rng.random([3,1])
W

Out[6]: array([[0.41848923],
[0.32338244],
[0.43007395]])

EVALUTING HYPOTHESIS(predicted y)

In [7]: #hpothesis
h = np.dot((W.T), x_Train)
h

Out[7]: array([[ 682.1053649 , 519.12061512, 777.82656714, 459.18817221,

972.28610511, 644.05292849, 497.77737408, 463.175453 ,
447.97647832, 484.84207648, 629.50071868, 648.47359113,
612.90152273, 1450.67542542, 411.75764504, 745.91839709,
428.14345796, 401.40940695, 845.84357106, 982.31096075,
573.12548261, 611.8246839 , 520.41414488, 636.61513236,
1259.66640278, 357.42939511, 473.20030864, 818.57275459,
713.15007914, 854.46820543, 595.97894434, 324.2310032 ,
661.83896269, 1016.15942544, 587.78438392]])

EVALUATING SQUARE ERROR

In [8]: #SE = square error = array of square errors of every indivisual data-point
SE = ((h - y_Train) ** 2)
print(SE)

[[1.59374927e+11 1.08491764e+11 1.35587569e+11 5.36111475e+10

2.90443081e+11 8.95541219e+10 9.88487576e+10 3.94164736e+10
4.47542587e+10 5.85713367e+10 5.72977572e+10 1.19959380e+11
1.08495202e+11 4.87831459e+11 6.73341479e+10 2.01739389e+11
8.96833929e+10 3.97996876e+10 2.49152875e+11 3.57625156e+11
6.36688516e+10 6.47133437e+10 5.87478636e+10 6.72175027e+10
3.27916952e+11 6.22714945e+10 2.15320871e+11 2.19193849e+11
2.24948016e+11 8.94282301e+10 1.22013299e+11 2.87559414e+10
9.87456219e+10 3.35106501e+11 8.14030604e+10]]

SUM OF SQUARE-ERROR

In [9]: #Sum_SE = sum of square errors

Sum_SE = np.sum(SE)
Sum_SE

Out[9]: 4917023281209.861

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 4/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

EVALUTING THE COST FUNCTION

In [10]: #Cost function

Cost_Func = (Sum_SE / (2 * s1))
Cost_Func

Out[10]: 70243189731.56944

GIVING THE VALUES OF HYPER-PARAMETERS

In [11]: #giving the no.of iterations

Iter_No = 60000
#giing the learning rate
Alpha = 0.00000000001

CREATING AN EMPTY LIST FOR COST-FUNCTIONS

In [12]: #Creating an empty list for cost functions

L = list([])
L.append(Cost_Func)

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 5/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

APPLYING GRADIENT DESCENT ALGORITHM

In [13]: print(f"Iteration 1:Cost funcion = {Cost_Func}")

loss = h - y_Train
#Evaluating the differtiables
Diff_W = (np.sum(loss * x_Train)) / s1
for i in range(2, Iter_No+1):
#Evaluating new / updated W(Thetas)
W -= (Alpha)*(Diff_W)
#new hypothesis
h1 = np.dot((W.T),x_Train)
#nes Square Error
SE = ((h1 - y_Train) ** 2)
Sum_SE = np.sum(SE)
#new updated cost funcion
Cost_Func = (Sum_SE / (2 * s1))
loss = h1 - y_Train
#new Differentiables
Diff_W = (np.sum(loss * x_Train)) / s1
#appending the new cost Function into the list
L.append(Cost_Func)
if (i % 6000 == 0):
print(f"Iteration {i}:Cost funcion = {Cost_Func}")

Iteration 1:Cost funcion = 70243189731.56944

Iteration 6000:Cost funcion = 40767959093.04384
Iteration 12000:Cost funcion = 24069605076.11786
Iteration 18000:Cost funcion = 14611694226.564194
Iteration 24000:Cost funcion = 9254754401.990297
Iteration 30000:Cost funcion = 6220595512.084834
Iteration 36000:Cost funcion = 4502054615.477992
Iteration 42000:Cost funcion = 3528676845.7321734
Iteration 48000:Cost funcion = 2977357825.9138227
Iteration 54000:Cost funcion = 2665091950.258117
Iteration 60000:Cost funcion = 2488225247.6887264

FINAL THETAS

In [14]: #Final Thethas

W

Out[14]: array([[159.96456673],
[159.86945994],
[159.97615145]])

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 6/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

PLOTTING THE COST-FUNCTION VS NO.OF ITERATIONS

GRAPH

In [15]: #I = no .of Iterations matrix

I = np.arange(1,60001)
fig = plt.figure()
fig.set_figheight(8)
fig.set_figwidth(15)
print(plt.xlabel("No.of Iterations"))
print(plt.ylabel("Cost Function"))
print(plt.plot(I, L, 'green', label = 'J vs Iter_No'))
print(plt.title("Cost function vs No of Iterations graph"))
print(plt.legend())

Text(0.5, 0, 'No.of Iterations')

Text(0, 0.5, 'Cost Function')
[<matplotlib.lines.Line2D object at 0x000002031541F850>]
Text(0.5, 1.0, 'Cost function vs No of Iterations graph')
Legend

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 7/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

VALIDATION

In [16]: h_Valid = np.dot(W.T,x_Train)

loss_Valid = h_Valid - y_Train
loss_Valid

Out[16]: array([[ -62894.76326182, -73468.97107238, 15326.59688089,

-5144.92785308, -59491.7510027 , 18240.74715659,
-69020.35542852, 29774.61235773, 9259.74774048,
-3015.13382618, 70947.62145922, -26621.18709574,
-27205.82768932, 16955.28694241, -56545.63177298,
-81400.37296181, -88392.39600747, -1661.45449111,
-82098.70983986, -113635.79774452, 30229.22873787,
47313.45723935, 14170.50676739, 54563.74957793,
48632.09219385, -73403.70104317, -230770.43438408,
-64529.85116647, -122647.29510743, 122315.65888705,
-55420.14629779, -9710.59934023, 12033.56745338,
-77749.61114216, 4423.46112675]])

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 8/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

VALIDATION GRAPH

In [17]: #Validation plot

h_Valid = np.reshape(h_Valid, newshape = ([35,1]))
y_Train = np.reshape(y_Train, newshape = ([35,1]))
print(plt.plot(h_Valid, y_Train, '.',label = 'Distortion in Validation'))
x = np.arange(750000)
y = x
print(plt.plot(x, y, '-',label = 'Y=X line'))
print(plt.ylabel("Predicted Values"))
print(plt.xlabel("Actual Values"))
print(plt.title("Validation Graph"))
plt.legend()

[<matplotlib.lines.Line2D object at 0x0000020316E284D0>]

[<matplotlib.lines.Line2D object at 0x0000020316E2B590>]
Text(0, 0.5, 'Predicted Values')
Text(0.5, 0, 'Actual Values')
Text(0.5, 1.0, 'Validation Graph')

Out[17]: <matplotlib.legend.Legend at 0x20315c5fb50>

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 9/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

TESTING

In [18]: #Testing
h_Test = np.dot(W.T,x_Test)
loss_Test = h_Test - y_Test
loss_Test

Out[18]: array([[ -19527.69304285, -31181.84611129, -3358.442232 ,

125649.64282632, 59437.64156625, -101997.30178757,
28527.74437035, 97184.77284259, -106516.75504901,
-43211.30326013, -3021.89101564, -46537.14666918]])

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 10/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

TESTING GRAPH

In [19]: #Testing plot

h_Test = np.reshape(h_Test, newshape = ([12,1]))
y_Test = np.reshape(y_Test, newshape = ([12,1]))
print(plt.plot(h_Test, y_Test, '.',label = 'Distortion in Testing'))
x = np.arange(750000)
y = x
print(plt.plot(x, y, '-',label = 'Y=X line'))
print(plt.ylabel("Predicted Values"))
print(plt.xlabel("Actual Values"))
print(plt.title("Testing Graph"))
plt.legend()

[<matplotlib.lines.Line2D object at 0x0000020313F1C9D0>]

[<matplotlib.lines.Line2D object at 0x0000020316EB32D0>]
Text(0, 0.5, 'Predicted Values')
Text(0.5, 0, 'Actual Values')
Text(0.5, 1.0, 'Testing Graph')

Out[19]: <matplotlib.legend.Legend at 0x20316e86410>

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 11/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

FINAL HYPER-PLANE and DATA-POINTS

In [20]: plt.axes(projection='3d')
plt.plot(Data_Set[:,0],Data_Set[:,1],Data_Set[:,2],'.',color = 'green')
plt.title('Data-Points')
ax = fig.add_subplot(111,projection='3d')
plt.xlabel('X1(Feature-1)')
plt.ylabel('X2(Feature-2)')
#Final Hyper-plane
a = W[1]
b = -1
c = W[2]
d = -(W[0])
x = np.linspace(-100,100,10)
y = np.linspace(-100,100,10)
X,Y = np.meshgrid(x,y)
Z = (d + a*X + b*Y) / c
fig = plt.figure()
ax = fig.add_subplot(111,projection='3d')
surf = ax.plot_surface(X, Z, Y , label = 'oooo')
ax.view_init(30,160)
plt.title('Final Hyper-Plane')
plt.xlabel('X1(Feature-1)')
plt.ylabel('X2(Feature-2)')

Out[20]: Text(0.5, 0.5, 'X2(Feature-2)')

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 12/13

11/2/23, 11:34 PM LINEAR REGRESSION WITH MULTIPLE VARIABLES - Jupyter Notebook

localhost:8890/notebooks/CAL201/SUPERVISED LEARNING/LINEAR REGRESSION WITH MULTIPLE VARIABLES.ipynb 13/13