Lesson 01 � Beginner

LINEAR REGRESSION:
SEEDHI LINE SE
PREDICTION.

Linear regression sabse basic aur important ML algorithm hai. Ek straight line se data fit karte hain � jaise ghar ka size se price predict karna, ya salary se monthly expense. Equation hai y = mx + b � bas itna hi.

WHY: Linear regression kyun seekhein?

Linear regression ML ki foundation hai � ye algorithm data ke beech relationship samajhta hai aur future predictions deta hai. House price prediction, salary estimation, sales forecasting � sab linear regression se shuru hota hai. Agar ek variable doosre ko affect kar raha hai, toh linear regression quantify karta hai ki kitna affect kar raha hai.

HOW: Linear regression kaam kaise karta hai

Linear regression ek simple equation dhundhta hai: y = mx + b. Algorithm sabse best line fit karta hai � wo line jo sabse kam error (distance) rakhe actual data points se. Ye Ordinary Least Squares (OLS) method se hota hai.

from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
import numpy as np

# House price prediction
X = np.array([[1000],[1500],[2000],[2500],[3000]]) # sq ft
y = np.array([50000,75000,100000,125000,150000]) # price

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
model = LinearRegression()
model.fit(X_train, y_train)

print(f"Coefficient: {model.coef_[0]:.2f}")
print(f"Intercept: {model.intercept_:.2f}")
print(f"R�: {model.score(X_test, y_test):.2f}")
print(f"Predict 2200sqft: ${model.predict([[2200]])[0]:.0f}")

FIT LINE: y = mx + b kya hai

Ye equation line ki kahani hai. m (coefficient) batata hai ki X ka 1 unit badhne se y kitna badhega. b (intercept) batata hai ki jab X = 0 ho, tab y ki value kya hai. Dono milke line define karte hain jo data ko best fit kare.

import numpy as np
from sklearn.linear_model import LinearRegression

# Salary prediction: experience (years) se salary predict karo
X = np.array([[1],[2],[3],[4],[5],[6],[7]])
y = np.array([30000,35000,45000,50000,60000,65000,75000])

model = LinearRegression()
model.fit(X, y)

m = model.coef_[0] # Slope
b = model.intercept_ # Intercept

print(f"Equation: y = {m:.0f}x + {b:.0f}")
print(f"Har saal ?{m:.0f} badhega salary")
print(f"Start mein ?{b:.0f} salary hogi")

# Prediction
years = 8
pred = model.predict([[years]])
print(f"\n{years} saal baad salary: ?{pred[0]:.0f}")

R�: Model kitna achha hai

R� (R-squared) batata hai ki model kitna data explain kar raha hai. 1.0 matlab perfect prediction, 0.0 matlab model ne kuch nahi samjha. Negative R� bhi ho sakta hai � matlab model mean se bhi bura hai!

from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error, r2_score
import numpy as np

# Multiple metrics
X = np.array([[1000],[1500],[2000],[2500],[3000]])
y = np.array([50000,75000,100000,125000,150000])

model = LinearRegression()
model.fit(X, y)

y_pred = model.predict(X)

# R� score
r2 = r2_score(y, y_pred)
# Mean Squared Error
mse = mean_squared_error(y, y_pred)

print(f"R�: {r2:.4f}")
print(f"MSE: {mse:.2f}")
print(f"RMSE: {np.sqrt(mse):.2f}")

# Perfect fit se compare
print(f"\nModel {r2*100:.1f}% data explain kar raha hai")

Try it: linear regression model build karo

Editor mein sample data diya hai � house prices predict karo. Coefficient aur R� check karo. Naya data daalo, model ko retrain karo aur dekho predictions kaise change hote hain.

Common linear regression mistakes

• Linear relationship assume karna: Har data linear nahi hota. Pehle scatter plot dekho � agar curve dikh raha hai toh polynomial regression ya transformation use karo, linear nahi.
• Outliers ka ignore karna: Ek extreme value poora model hilaa sakta hai. Pehle data clean karo, outliers handle karo.
• R� pe blindly bharosa karna: R� high ho toh bhi check karo ki model meaningful hai ya nahi. Correlation ✓ causation hai hamesha.
• Train test split na karna: Sab data pe train karoge toh overfitting pakka hai. Hamesha 20-30% data test ke liye chhodo.
• Feature scaling na karna: Multiple features ho toh scaling zaroori hai � nahi toh ek feature dominating kar sakta hai. StandardScaler ya MinMaxScaler use karo.