This project is an end-to-end Machine Learning system that predicts house prices using the Boston Housing dataset. It covers the complete ML workflow starting from data preprocessing to deployment through an interactive web application.
The goal of this project is not just to train a model, but to build a practical system that can be used to estimate property prices based on different input features.
Accurately estimating house prices is important for buyers, sellers, and real estate professionals. Manual estimation is often inconsistent and depends heavily on experience.
This project uses machine learning to provide a data-driven approach for predicting house prices based on various factors such as number of rooms, crime rate, and population characteristics.
- Python
- Pandas & NumPy
- Scikit-learn
- Streamlit
- Matplotlib
The project follows a structured machine learning pipeline:
-
Data Preprocessing
- Loaded dataset and handled inconsistencies
- Cleaned column names
- Checked for missing values
-
Feature Engineering
- Created new meaningful features like:
- Rooms per age
- Tax per room
- Improved model performance using derived features
- Created new meaningful features like:
-
Model Training
- Trained multiple models:
- Linear Regression
- Random Forest Regressor
- Selected the best-performing model
- Trained multiple models:
-
Model Evaluation
- Evaluated using:
- RMSE (Root Mean Squared Error)
- R² Score
- Compared model performance
- Evaluated using:
-
Model Deployment
- Built an interactive UI using Streamlit
- Users can input feature values and get instant predictions
Some of the key features include:
- RM → Average number of rooms
- LSTAT → Percentage of lower income population
- PTRATIO → Student-teacher ratio
- CRIM → Crime rate
- TAX → Property tax rate
- and other numerical features from the dataset
Example input:
- RM: 6
- LSTAT: 12
- PTRATIO: 18
- CRIM: 1
- Other features set to default values
Output:
- Predicted House Price: ~ $22,942.00
- Number of rooms (RM) has a strong positive impact on price
- Higher LSTAT values tend to reduce house prices
- Crime rate and tax also influence predictions significantly
The Streamlit app allows users to:
- Adjust property features using sliders
- Predict house price instantly
- View feature importance graph
This system can be useful for:
- Real estate agents to estimate property prices
- Buyers to evaluate if a property is fairly priced
- Investors to identify undervalued properties
- Property platforms to automate price estimation
pip install -r requirements.txt
python main.py
cd app
streamlit run app.py
house-price-prediction-model/
│
├── app/ # Streamlit app
├── data/ # Dataset
├── models/ # Saved model
├── src/ # ML pipeline code
│
├── main.py # Main execution file
├── image-1.png # Screenshot of the project
├── image-2.png # Screenshot of the project
├── image.png # Screenshot of the project
├── README.md
└── requirements.txt
- Building an end-to-end ML pipeline
- Importance of feature engineering
- Handling real-world data issues
- Deploying ML models as applications
- Improving user interaction with simple UI
- Add XGBoost for better performance
- Improve UI design for better user experience
- Add real-time data input integration
This project demonstrates how machine learning can be used to solve real-world problems by combining data processing, model building, and deployment into a single system. It reflects both technical understanding and practical implementation skills.