Linear Regression implementaion

AI Discussions
,
1

I am stuck in linear regression. When I inserted 2 to 3 values on my own in X and Y, it worked well. But when I trained the dataset from Kaggle, it gave w, b, and cost function values as NaN. Could somebody please guide me on what I am doing wrong? I have attached the dataset and provided the code.

from google.colab import files
uploaded = files.upload()
import pandas as pd

# Read the uploaded file
df = pd.read_csv('Study_vs_Score_data.csv')

# View top rows to confirm
print(df.head())
X = df['Attendance_Hours']
y = df['Final_Marks']

print(X.head())
print(y.head())
# Drop rows with NaN
df = df.dropna()

# Convert columns to numeric
df['Attendance_Hours'] = pd.to_numeric(df['Attendance_Hours'], errors='coerce')
df['Final_Marks'] = pd.to_numeric(df['Final_Marks'], errors='coerce')

# Drop any rows that still have NaN after conversion
df = df.dropna()
# Import required libraries
import numpy as np
import matplotlib.pyplot as plt

# Step 1: Prepare the data
# Let's create simple data for example
# X: input feature, y: target output

X = df['Attendance_Hours']
y = df['Final_Marks']  # Here, y = 2 * X

# Number of data points
m = len(X)
print(f"Number of rows are: {m}")
# Step 2: Initialize parameters (weight and bias)
w = 0  # initial slope
b = 0  # initial intercept

# Step 3: Set learning rate and number of iterations
learning_rate = 0.01
iterations = 1000

# Step 4: Perform Gradient Descent
for i in range(iterations):
    # Calculate the predicted value
    y_pred = w * X + b
    
    # Calculate cost (mean squared error)
    cost = (1/m) * sum((y_pred - y) ** 2)
    
    # Calculate gradients
    dw = (2/m) * sum((y_pred - y) * X)
    db = (2/m) * sum(y_pred - y)
    
    # Update parameters
    w = w - learning_rate * dw
    b = b - learning_rate * db
    
    # Print cost every 100 iterations
    if i % 100 == 0:
        print(f"Iteration {i}, Learning_rate: {learning_rate}, Cost: {cost:.4f}, w: {w:.4f}, b: {b:.4f}")

# Step 5: Print final results
print(f"\nFinal weight (w): {w}")
print(f"Final bias (b): {b}")

# Step 6: Plot the data and regression line
plt.scatter(X, y, color='blue', label='Actual Data')
plt.plot(X, w*X + b, color='red', label='Regression Line')
plt.xlabel('X')
plt.ylabel('y')
plt.legend()
plt.show()
[Study_vs_Score_data.csv|attachment](upload://gdslqmRLWmDcXP2IdBTMo7CEd04.csv) (35.5 KB)
[Study_vs_Score_data.csv|attachment](upload://gdslqmRLWmDcXP2IdBTMo7CEd04.csv) (35.5 KB)
2

Welcome to the community.
I think this requires information on the dataset you are working on. Also, you could consider using a kaggle notebook so you can interact directly on the platform without needing to download and then upload to colab.

1 Like
3

I believe the cost should divide by 2m, and the gradients should not be multiplied by 2.

Is this for a course assignment?

1 Like
4

I am a beginner, so I don’t know how to interact with Kaggle notebooks. I also attached the dataset with the query, but I don’t know why it is not visible to you. It was the Study vs Score dataset. I found the mistake I was making: I was using it directly without data cleaning and standardization. But when I cleaned and standardized it, it worked perfectly. Thank you for your kind information.

2 Likes
5

Thank you very much for your guidance. And yes, I am doing a Machine Learning Specialization course from DeepLearning.AI on Coursera, and I have just completed the first module of the first course.

1 Like
6

It was not an assignment. It was Optional Lab practice.

7

Thanks for the update.