House Price Prediction in Python

What we are building

A house price predictor is a machine learning model that estimates a price from features like size and number of rooms. Because the answer is a number, this is a regression problem, and we solve it with linear regression using scikit-learn.

Regression, not classification

Spam detection sorts messages into categories — that is classification. House price prediction gives a continuous number like Rs 42,50,000 — that is regression. Linear regression learns a relationship between the inputs (features) and the output number, then uses it to predict prices for new houses.

The data (features and price)

# feature: area in sq ft   ->  target: price (in lakh)
X = [[500], [800], [1000], [1500]]   # areas
y = [25, 40, 50, 75]                 # prices

Each house has features (here just area) and a known price. The model learns how area relates to price. Real datasets add bedrooms, location, age and more.

Training the model

from sklearn.linear_model import LinearRegression

model = LinearRegression()
model.fit(X, y)     # learn the relationship

LinearRegression finds the best line through the data. fit(X, y) trains it, so it learns roughly how much price increases per square foot.

Predicting a price

new_house = [[1200]]           # 1200 sq ft
price = model.predict(new_house)
print("Predicted price:", price[0], "lakh")

We pass the new house's features to predict, and it returns the estimated price based on the learned line. More features give more accurate predictions.

Checking the model

from sklearn.metrics import r2_score

pred = model.predict(X)
print("R2 score:", r2_score(y, pred))   # closer to 1 is better

To trust the model, measure its accuracy. r2_score close to 1 means the line fits the data well. In practice, always test on data the model has not seen.

Common mistakes

• Using classification for a number. Price is continuous — use regression, not a classifier.
• Too few features. One feature is limited; add rooms, location, age for better results.
• Not scaling features. Very different ranges (area vs rooms) can help from scaling.
• Testing on training data only. Always check accuracy on unseen houses.

Summary

• Predicting a price is regression (a number), not classification.
• Linear regression learns a line between features and price.
• Train with model.fit(X, y), predict with model.predict().
• Add more features for more accurate predictions.
• Check accuracy (like R-squared) on unseen test data.

Next, compare with classification in spam detection, or learn ML types in supervised learning.