House Price Prediction

🏠 House Price Prediction in Ames, Iowa

Predicting real estate prices using advanced regression techniques

“Accurate home valuation — powered by data.”

📎 Full Analysis:
👉 View Jupyter Notebook on GitHub

📎 Competition Source:
👉 Kaggle: House Prices - Advanced Regression Techniques

House Price Prediction

📌 One-Line Summary

This project predicts house prices in Ames, Iowa using 81 property features such as area, location, condition, and building type.
It applies data preprocessing, feature engineering, and regression modeling to estimate realistic sale prices.

1️⃣ How It Was Built

1. Data Collection

Train Data: 1,460 records, 81 features
Test Data: 1,459 records, 80 features (SalePrice excluded)
Total: 2,919 property listings
Data from Kaggle competition dataset

2. Data Preparation

Checked missing values (e.g., LotFrontage, MasVnrArea)
Filled missing values using median or mode depending on the feature type
Removed outliers (e.g., extreme GrLivArea values)
Applied log transformation to SalePrice to normalize skewed distribution
Converted categorical features into numerical form using one-hot encoding

3. Exploratory Data Analysis (EDA)

Numerical features: Plotted scatter graphs with SalePrice to detect trends
→ Found GrLivArea has strong linear correlation with price
Categorical features: Compared median prices across categories (e.g., neighborhood)
Observed that larger living area and better overall quality strongly increase house price

4. Feature Engineering

Created new features such as:
- TotalSF = Total square footage (basement + 1st floor + 2nd floor)
- Age = Years since construction
Dropped low-impact or highly correlated redundant features

5. Model Training (OLS Regression)

Used Ordinary Least Squares (OLS) regression to predict prices
Checked multicollinearity using Variance Inflation Factor (VIF)
Selected final set of features after removing high-VIF variables

📊 Evaluation Metric:

Root Mean Squared Error (RMSE) used for performance check
Final RMSE (log-transformed target): ~0.12 on validation set

6. Results

Feature	Impact on Price
`OverallQual`	Very High
`GrLivArea`	High
`GarageCars`	High
`TotalSF`	High
`Neighborhood`	Moderate

Example Prediction:

House: 2-story, built in 2005, 2,000 sqft, good neighborhood
Predicted Price: ~$197,500

2️⃣ Real-World Use

The OLS model can be used by real estate agencies to estimate property prices
Buyers & sellers can check if a listing is fairly priced
Government agencies can use it for property tax assessment

🛠 Technologies Used

Step	Technology
Data Storage	CSV (Kaggle Dataset)
Model Dev	Python, Pandas, NumPy, statsmodels
Visualization	Matplotlib, Seaborn
Environment	Jupyter Notebook

💡 Key Learnings

Log-transforming skewed price data improves model performance
Removing multicollinear features increases stability of regression coefficients
Even simple linear models can perform well with good feature engineering

🔗 GitHub Repository

📂 View Project on GitHub