House Price Prediction Model

Overview

This project implements a machine learning model to predict house prices using various features such as year built, number of bedrooms, amenities, and location data. The model uses Random Forest Regression and compares its performance with OLS (Ordinary Least Squares) and Decision Tree approaches.

Data Description

The dataset contains housing data from 2016-2017 with the following key features:

• Property characteristics (bedrooms, bathrooms, square footage)
• Building information (year built, number of floors)
• Financial details (maintenance costs, taxes, common charges)
• Amenities (garage, pets allowed, fuel type)
• Property type (coop/condo)
• Kitchen and dining room specifications

Technical Implementation

Data Preprocessing

• Feature selection reducing to 25 key features
• Binary variable conversion to 0/1 dummies
• Min-Max scaling for year built feature
• Missing value imputation using Random Forest
• Categorical variable encoding

Models Implemented

1. Random Forest Regressor (Primary Model)
2. OLS (Ordinary Least Squares) Regression
3. Decision Tree Regressor (for comparison)

Key Features

• Automated missing value handling
• Feature importance analysis
• Model performance comparison
• Visualization of decision trees
• Price prediction functionality for new properties

Model Performance

The project includes various performance metrics:

• R-squared score
• Root Mean Squared Error (RMSE)
• Training and testing set evaluations
• Feature importance rankings

Requirements

- pandas
- numpy
- statsmodels
- scikit-learn
- matplotlib
- graphviz
- PIL

Usage

1. Load and preprocess the data:

data_frame = pd.read_csv("housing_data_2016_2017.csv")

2. Train the model:

rf_model = RandomForestRegressor(n_estimators=100, random_state=42)
rf_model.fit(X_train, y_train)

3. Make predictions:

predicted_price = rf_model.predict(new_house_df)

File Structure

• main.ipynb: Main Jupyter notebook containing all analysis and model implementation
  • Data cleaning and preprocessing
  • Model training and evaluation
  • Visualization and analysis
  • Prediction functionality

Future Improvements

• Feature engineering optimization
• Hyperparameter tuning
• Additional model architectures
• Cross-validation implementation
• Enhanced visualization features

Contributors

This project is maintained by Carlos Vega.