Linear regression with gradient descent to predict house prices using area one var in python

Question

I am trying to train a model using linear regression and gradient descent to predict the house prices based on area in sq ft but idk what is  wrong, the predicted hx (hypothesis) is larger than values. have not coded the for the prediction as i am getting error. help me out. ty

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

db = pd.read_csv("train.csv")
x = np.array(db.iloc[:40,5])
y = np.array(db.iloc[:40,11])
m = len(x)
t0=0
t1=1
alpha = 0.001
for i in range(m):
    hx = t0 + t1*x
    dt0 = (-2/m)*sum((hx - y))
    dt1 = (-2/m)*sum(x*(hx - y))
    t0 = t0 - dt0*alpha
    t1 = t1 - dt1*alpha
print(t0,t1)
plt.scatter(x,y)
plt.plot(x,hx)

also dataset,

x = [1300.236407 , 1275. , 933.1597222, 929.9211427, 999.009247 , 1250. , 1495.053957 , 1181.012946 , 1040. , 879.1208791, 1350.308642 , 1333.010179 , 927.1779023, 1122.171946 , 649.9837504, 1394.117647 , 1800.08471 , 2124.896706 , 1100. , 2178.649237 , 881.1435285, 944.8818898, 1310.147689 , 630.00063 , 1219.80971 , 780.141844 , 1600. , 1180.412371 , 1000. , 1000. , 1400.107701 , 943.1266076, 1150.146382 , 864.0674394, 857.7861968, 1174.210077 , 1020.087884 , 1650.165017 , 1000. , 1300.052002 ]

y = [ 55. , 51. , 43. , 62.5, 60.5, 42. , 66.5, 52. , 41.6, 36. , 35. , 110. , 48. , 62. , 20. , 71.1, 85. , 180. , 22. , 120. , 45. , 42. , 55. , 300. , 50. , 27.5, 46. , 22.9, 39. , 12.5, 52. , 33. , 55. , 82. , 240. , 55. , 65. , 65. , 35. , 75. ]

Answer 1

Apart from instructional purposes, I'm not sure why you require gradient descent for this and why you need to develop an algorithm from the start. For this, I propose using well-known tools such as Sklearn or Statsmodels.

Otherwise, linear regression can be performed directly like this, or you can utilize the aforementioned programs to choose which solver to use if you wish to use logistic (a type of linear regression) (I am not sure whether gradient descent is one of them as there are more controllable ways to solve linear problems that that).