33. Linear Regression in Python with sklearn
33.5 Common Linear Regression Mistakes
Why. Beginners often get a running notebook that is still wrong for business. Spot these traps early.
What. Common linear-regression mistakes: fitting on all rows with no test set; putting the label inside X; expecting a perfect line on messy data; forgetting units (₹ vs lakhs); using linear regression for a clear yes/no target.
How.
Mistake checklist
- No split. Always hold test rows (Chapter 31).
- Leakage. Do not include future rent or “final_price_paid” as a feature when predicting rent.
- Wrong problem type. Churn yes/no → logistic (Chapter 34), not LinearRegression.
- One wild outlier. A ₹9 lakh typo rent can tilt the line — clean first (Chapter 9 habit).
- Over-trust R² on 10 rows. Tiny samples look perfect; they are not proof.
# BAD: label inside features
# X_bad = df[["area_sqft", "rent_inr"]] # never do this
# GOOD
X_good = df[["area_sqft"]]
y_good = df["rent_inr"]
assert "rent_inr" not in X_good.columns
Ravindra Bagale's Tip
Model fit jhala mhanje project complete nahi. Test predictions + business sense = complete.