🏠 Singapore Resale Flat Prices Predicting

This project aims to develop a machine learning model and deploy it as a user-friendly web application to predict the resale prices of flats in Singapore. The predictive model is trained using historical resale flat transaction data and is designed to assist both potential buyers and sellers in estimating the market value of a flat based on key features.

📌 Project Objective

To build and deploy a machine learning model that:

• Analyzes historical HDB resale flat data
• Predicts the resale value of a flat based on user input
• Offers a web interface using Streamlit for ease of use

🛠️ Tech Stack

• Python
• Pandas & NumPy – Data preprocessing
• Scikit-learn – Model building and evaluation
• Streamlit – Web app interface
• VSCode – Project development environment

📚 Data Source

The dataset was obtained from the official Singapore Housing Development Board (HDB) via data.gov.sg.
Five different datasets were provided by the team GUVI and downloaded for this project.

🧠 Project Workflow

1. Environment Setup

• Project developed in VSCode
• A dedicated virtual environment was created
• Required packages were installed using pip

2. Data Collection & Preprocessing

• Merged five datasets covering different timelines
• Performed:
  • Data cleaning
  • Handling of missing values
  • Standardization of column names

3. Feature Selection

• Selected key features for prediction:
  • floor_area_sqm
  • flat_type
  • town
  • storey_range
  • flat_model
  • age_of_flat
  • remaining_lease_months

4. Feature Engineering

• Applied OneHotEncoding to categorical columns:
  • flat_type, town, storey_range, flat_model
• Added encoded features back into the dataset with descriptive column names

5. Model Selection & Training

• Tried multiple regression models: Linear Regression, Decision Trees, and Random Forest
• Random Forest Regressor selected as the final model due to superior performance
• Trained the model on encoded categorical + numerical features
  • Features: Town, Storey Range, Flat Model, Flat Type, Floor Area (sqm), Remaining Lease (months), Age of Flat (years)
  • Target: Resale Price

6. Model Evaluation

• Evaluated using:
  • Mean Absolute Error (MAE)
  • Root Mean Squared Error (RMSE)
  • R² Score
• Random Forest model showed:
  • Lowest MAE and RMSE
  • Highest R² score, indicating strong predictive accuracy

7. Streamlit Web App Development

• Built a Streamlit interface to accept 7 user inputs:
  • Town, Storey Range, Flat Model, Flat Type, Floor Area (sqm), Remaining Lease (months), Age of Flat (years)
• Inputs are mapped to their encoded formats and passed to the trained model
• Predictions are made upon clicking the "Predict Resale Price" button

🖥️ Key Features of the Web App

• Interactive form to input flat details
• Real-time prediction using trained ML model
• Clean and intuitive UI built with Streamlit
• Fully functional and tested on live environment

✅ Results

• The model provides accurate price estimations for resale flats in Singapore.
• Helps buyers make informed decisions and helps sellers estimate property value.
• Demonstrates end-to-end ML pipeline from data preprocessing to model deployment.

🚀 Future Improvements

• Integrate model retraining with updated datasets
• Add data visualizations for trends by town/flat type
• Expand deployment to cloud platforms with persistent storage

📄 Deliverables

• ✅ Trained ML model
• ✅ Streamlit web application
• ✅ Encoders and preprocessing pipeline
• ✅ Deployment-ready app with documentation