Scribd
Upload
13 views

Nikitha

This document contains code to perform simple linear regression on a housing price dataset to predict prices based on square meters. It loads and explores the Paris housing dataset, selects relevant features correlated with price, then uses gradient descent with different learning rates to fit a linear regression model and analyze the cost function over iterations. Gradient descent with a learning rate of 0.1 converges faster than with a learning rate of 0.01 based on the plotted cost histories.

Uploaded by

Chakri Chakradhar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
13 views

Nikitha

This document contains code to perform simple linear regression on a housing price dataset to predict prices based on square meters. It loads and explores the Paris housing dataset, selects relevant features correlated with price, then uses gradient descent with different learning rates to fit a linear regression model and analyze the cost function over iterations. Gradient descent with a learning rate of 0.1 converges faster than with a learning rate of 0.01 based on the plotted cost histories.

Uploaded by

Chakri Chakradhar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

import pandas as pd

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression

dataset = pd.read_csv("/content/ParisHousing.csv")
dataset.head(20)

squareMeters numberOfRooms hasYard hasPool floors cityCode \


0 75523 3 0 1 63 9373
1 80771 39 1 1 98 39381
2 55712 58 0 1 19 34457
3 32316 47 0 0 6 27939
4 70429 19 1 1 90 38045
5 39223 36 0 1 17 39489
6 58682 10 1 1 99 6450
7 86929 100 1 0 11 98155
8 51522 3 0 0 61 9047
9 39686 42 0 0 15 71019
10 23563 21 0 1 90 91058
11 96470 74 1 0 21 92029
12 19127 31 1 0 5 7475
13 13087 44 1 0 77 40475
14 79770 3 0 1 69 54812
15 75985 60 1 0 67 6517
16 64169 88 0 1 6 61711
17 99371 31 1 1 16 96297
18 25966 37 1 1 17 22818
19 41792 43 1 1 10 80768

cityPartRange numPrevOwners made isNewBuilt hasStormProtector


\
0 3 8 2005 0 1

1 8 6 2015 1 0

2 6 8 2021 0 0

3 10 4 2012 0 1

4 3 7 1990 1 0

5 8 6 2012 0 1

6 10 9 1995 1 1

7 3 4 2003 1 0

8 8 3 2012 1 1
9 5 8 2021 1 1

10 6 8 1993 1 0

11 4 2 2011 1 1

12 2 9 2008 0 0

13 8 4 2004 1 0

14 10 5 2018 0 1

15 6 9 2009 1 1

16 3 9 2011 1 1

17 7 8 2013 1 1

18 3 1 2016 0 0

19 9 5 2017 1 1

basement attic garage hasStorageRoom hasGuestRoom price


0 4313 9005 956 0 7 7559081.5
1 3653 2436 128 1 2 8085989.5
2 2937 8852 135 1 9 5574642.1
3 659 7141 359 0 3 3232561.2
4 8435 2429 292 1 4 7055052.0
5 2009 4552 757 0 1 3926647.2
6 5930 9453 848 0 5 5876376.5
7 6326 4748 654 0 10 8696869.3
8 632 5792 807 1 5 5154055.2
9 5198 5342 591 1 3 3970892.1
10 703 852 684 1 10 2366397.3
11 5414 1172 716 1 9 9652258.1
12 5387 4430 374 0 4 1914688.8
13 1745 724 582 0 0 1320803.4
14 8871 7117 240 0 7 7986665.8
15 4878 281 384 1 5 7607322.9
16 3054 129 726 0 9 6420823.1
17 3258 6296 354 1 8 9944705.3
18 8257 2557 162 0 6 2604486.6
19 2950 9573 572 1 5 4187667.7

dataset.shape

(10000, 17)

dataset.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10000 entries, 0 to 9999
Data columns (total 17 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 squareMeters 10000 non-null int64
1 numberOfRooms 10000 non-null int64
2 hasYard 10000 non-null int64
3 hasPool 10000 non-null int64
4 floors 10000 non-null int64
5 cityCode 10000 non-null int64
6 cityPartRange 10000 non-null int64
7 numPrevOwners 10000 non-null int64
8 made 10000 non-null int64
9 isNewBuilt 10000 non-null int64
10 hasStormProtector 10000 non-null int64
11 basement 10000 non-null int64
12 attic 10000 non-null int64
13 garage 10000 non-null int64
14 hasStorageRoom 10000 non-null int64
15 hasGuestRoom 10000 non-null int64
16 price 10000 non-null float64
dtypes: float64(1), int64(16)
memory usage: 1.3 MB

corr = dataset.corr()
plt.figure(figsize=(10, 8))
sns.heatmap(corr, annot=True, cmap='coolwarm', fmt=".2f",
linewidths=0.5)
plt.title("Correlation Heatmap")
plt.show()
correlation_with_price = dataset.corr()['price'].abs()

threshold = 0.01

highly_correlated_columns =
correlation_with_price[correlation_with_price >
threshold].index.tolist()

print("Columns highly correlated with 'Price':")


print(highly_correlated_columns)

Columns highly correlated with 'Price':


['squareMeters', 'numPrevOwners', 'isNewBuilt', 'garage', 'price']

df = dataset[highly_correlated_columns]
df.head()

squareMeters numPrevOwners isNewBuilt garage price


0 75523 8 0 956 7559081.5
1 80771 6 1 128 8085989.5
2 55712 8 0 135 5574642.1
3 32316 4 0 359 3232561.2
4 70429 7 1 292 7055052.0

Gradient Desent, Learning rate, Cost Function

X = df['squareMeters'].values
y = df['price'].values

learning_rate = 0.01
iterations = 10

# Initialize coefficients (slope and intercept)


b0 = 0 # Intercept
b1 = 0 # Slope

# Lists to store the history of coefficients and cost


b0_history = []
b1_history = []
cost_history = []

# Gradient Descent
for iteration in range(iterations):
# Calculate predictions
y_pred = b0 + b1 * X

# Calculate the cost (mean squared error)


cost = np.mean((y_pred - y) ** 2)

# Calculate gradients
gradient_b0 = np.mean(y_pred - y)
gradient_b1 = np.mean((y_pred - y) * X)

# Update coefficients using gradients and learning rate


b0 -= learning_rate * gradient_b0
b1 -= learning_rate * gradient_b1

# Append coefficients and cost to history lists for visualization


b0_history.append(b0)
b1_history.append(b1)
cost_history.append(cost)

# Plot the cost history


plt.figure(figsize=(10, 4))
plt.subplot(1, 2, 1)
plt.plot(cost_history)
plt.xlabel('Iterations')
plt.ylabel('Cost')
plt.title('Cost History')

plt.tight_layout()
plt.show()

# Print the final coefficients and cost


print("Final Intercept (b0):", b0)
print("Final Slope (b1):", b1)
print("Final Cost:", cost_history[-1])

Final Intercept (b0): -2.4127252943307258e+72


Final Slope (b1): -1.6037599090064153e+77
Final Cost: 7.759017749802741e+148

X = df['squareMeters'].values
y = df['price'].values

learning_rate = 0.1
iterations = 10

# Initialize coefficients (slope and intercept)


b0 = 0 # Intercept
b1 = 0 # Slope
# Lists to store the history of coefficients and cost
b0_history = []
b1_history = []
cost_history = []

# Gradient Descent
for iteration in range(iterations):
# Calculate predictions
y_pred = b0 + b1 * X

# Calculate the cost (mean squared error)


cost = np.mean((y_pred - y) ** 2)

# Calculate gradients
gradient_b0 = np.mean(y_pred - y)
gradient_b1 = np.mean((y_pred - y) * X)

# Update coefficients using gradients and learning rate


b0 -= learning_rate * gradient_b0
b1 -= learning_rate * gradient_b1

# Append coefficients and cost to history lists for visualization


b0_history.append(b0)
b1_history.append(b1)
cost_history.append(cost)

# Plot the cost history


plt.figure(figsize=(10, 4))
plt.subplot(1, 2, 1)
plt.plot(cost_history)
plt.xlabel('Iterations')
plt.ylabel('Cost')
plt.title('Cost History')

plt.tight_layout()
plt.show()

# Print the final coefficients and cost


print("Final Intercept (b0):", b0)
print("Final Slope (b1):", b1)
print("Final Cost:", cost_history[-1])
Final Intercept (b0): -2.4127259493867817e+82
Final Slope (b1): -1.603760344427987e+87
Final Cost: 7.759021541641718e+166

from sklearn.model_selection import train_test_split

X = df['squareMeters'].values
y = df['price'].values
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=0.25, random_state=42)

from sklearn.linear_model import LinearRegression

model = LinearRegression()

X_train = X_train.reshape(-1, 1)
X_test = X_test.reshape(-1, 1)
# Fit the model to the data
model.fit(X_train, y_train)

# Make predictions
y_pred = model.predict(X_test)

from sklearn.metrics import r2_score, mean_squared_error

r2 = r2_score(y_test, y_pred)
mse = mean_squared_error(y_test, y_pred)
print("R-squared:", r2)
print("Mean Squared Error:", mse)

# Plot each independent variable against the dependent variable


for i in range(1):
plt.figure(figsize=(6, 4))
plt.scatter(X_train[:, i], y_train, label='Data')
plt.xlabel(df.columns[i])
plt.ylabel('Price')
plt.title(f'Scatter Plot of {df.columns[i]} vs. Price')

# Plot the regression line


sorted_indices = np.argsort(X_test[:, i])
plt.plot(X_test[:, i][sorted_indices], y_pred[sorted_indices],
color='red', label='Linear Regression')

plt.legend()
plt.show()

R-squared: 0.999998793097589
Mean Squared Error: 10440151.787275104

Gradient Desent, Cost Function, Learning Rate for Multi Regression

X = df[['squareMeters', 'numPrevOwners', 'isNewBuilt', 'garage']]


y = df['price']
def normalize(feature):
"""Standardize the feature using Z-score normalization."""
return (feature - np.mean(feature)) / np.std(feature)

# Hyperparameters
alpha = 0.01
num_iterations = 10

# Initialization
m = len(X['squareMeters'])
X0 = np.ones(m)
X1 = normalize(np.array(X['isNewBuilt']))
X2 = normalize(np.array(X['numPrevOwners']))
X3 = normalize(np.array(X['squareMeters']))
X4 = normalize(np.array(X['garage']))
y = normalize(np.array(y))

X = np.array([X0, X1, X2, X3, X4]).T


theta = np.zeros(5)

# Gradient Descent
for _ in range(num_iterations):
y_pred = np.dot(X, theta)
gradient = (1/m) * np.dot(X.T, (y_pred - y))
theta -= alpha * gradient

print("Parameters:", theta)

Parameters: [-8.27782287e-19 -9.70733553e-04 1.51884192e-03


9.56150111e-02
-1.57511062e-03]

X = df[['squareMeters', 'numPrevOwners', 'isNewBuilt', 'garage']]


y = df['price']

def normalize(feature):
"""Standardize the feature using Z-score normalization."""
return (feature - np.mean(feature)) / np.std(feature)

# Hyperparameters
alpha = 0.10
num_iterations = 10

# Initialization
m = len(X['squareMeters'])
X0 = np.ones(m)
X1 = normalize(np.array(X['isNewBuilt']))
X2 = normalize(np.array(X['numPrevOwners']))
X3 = normalize(np.array(X['squareMeters']))
X4 = normalize(np.array(X['garage']))
y = normalize(np.array(y))

X = np.array([X0, X1, X2, X3, X4]).T


theta = np.zeros(5)

# Gradient Descent
for _ in range(num_iterations):
y_pred = np.dot(X, theta)
gradient = (1/m) * np.dot(X.T, (y_pred - y))
theta -= alpha * gradient

print("Parameters:", theta)

Parameters: [ 2.11741735e-17 -4.05175641e-03 6.47160620e-03


6.51187869e-01
-6.73095587e-03]

X = df[['squareMeters', 'numPrevOwners', 'isNewBuilt', 'garage']]


y = df['price']

def normalize(feature):
"""Standardize the feature using Z-score normalization."""
return (feature - np.mean(feature)) / np.std(feature)

# Hyperparameters
alpha = 0.01
num_iterations = 20

# Initialization
m = len(X['squareMeters'])
X0 = np.ones(m)
X1 = normalize(np.array(X['isNewBuilt']))
X2 = normalize(np.array(X['numPrevOwners']))
X3 = normalize(np.array(X['squareMeters']))
X4 = normalize(np.array(X['garage']))
y = normalize(np.array(y))

X = np.array([X0, X1, X2,X3,X4]).T


theta = np.zeros(5)
# Cost history to store MSE values for each iteration
cost_history = []

# Gradient Descent
for _ in range(num_iterations):
y_pred = np.dot(X, theta)
cost = (1/m) * sum((y_pred - y)**2)
cost_history.append(cost)

gradient = (1/m) * np.dot(X.T, (y_pred - y))


theta -= alpha * gradient

plt.plot(cost_history)
plt.title('Cost Function Over Iterations (Multiple Regression)')
plt.xlabel('Iterations')
plt.ylabel('Cost (MSE)')
plt.show()

X = df[['squareMeters', 'numPrevOwners', 'isNewBuilt', 'garage']]


y = df['price']
def normalize(feature):
"""Standardize the feature using Z-score normalization."""
return (feature - np.mean(feature)) / np.std(feature)

# Hyperparameters
alpha = 0.10
num_iterations = 20

# Initialization
m = len(X['squareMeters'])
X0 = np.ones(m)
X1 = normalize(np.array(X['isNewBuilt']))
X2 = normalize(np.array(X['numPrevOwners']))
X3 = normalize(np.array(X['squareMeters']))
X4 = normalize(np.array(X['garage']))
y = normalize(np.array(y))

X = np.array([X0, X1, X2,X3,X4]).T


theta = np.zeros(5)

# Cost history to store MSE values for each iteration


cost_history = []

# Gradient Descent
for _ in range(num_iterations):
y_pred = np.dot(X, theta)
cost = (1/m) * sum((y_pred - y)**2)
cost_history.append(cost)

gradient = (1/m) * np.dot(X.T, (y_pred - y))


theta -= alpha * gradient

plt.plot(cost_history)
plt.title('Cost Function Over Iterations (Multiple Regression)')
plt.xlabel('Iterations')
plt.ylabel('Cost (MSE)')
plt.show()
from sklearn.model_selection import train_test_split

X = df[['squareMeters', 'numPrevOwners', 'isNewBuilt',


'garage']].values
y = df['price'].values

X_train, X_test, y_train, y_test = train_test_split(X, y,


test_size=0.25, random_state=42)

from sklearn.linear_model import LinearRegression

model = LinearRegression()

# Fit the model to the data


model.fit(X_train, y_train)

# Make predictions
y_pred = model.predict(X_test)

from sklearn.metrics import r2_score, mean_squared_error

r2 = r2_score(y_test, y_pred)
mse = mean_squared_error(y_test, y_pred)

print("R-squared:", r2)
print("Mean Squared Error:", mse)

# Plot each independent variable against the dependent variable


for i in range(1):
plt.figure(figsize=(6, 4))
plt.scatter(X_train[:, i], y_train, label='Data')
plt.xlabel(df.columns[i])
plt.ylabel('Price')
plt.title(f'Scatter Plot of {df.columns[i]} vs. Price')

# Plot the regression line


sorted_indices = np.argsort(X_test[:, i])
plt.plot(X_test[:, i][sorted_indices], y_pred[sorted_indices],
color='red', label='Linear Regression')

plt.legend()
plt.show()

R-squared: 0.9999987942704442
Mean Squared Error: 10430006.155390566

You might also like