New York Housing Market: End-to-End Price Prediction

Project Overview

An end-to-end Machine Learning solution designed to estimate real estate prices in New York City. The project moves from raw data scraping/collection to a fully functional interactive web application, helping users understand market trends and property values.

Key Performance Indicators (KPIs)

• Model Accuracy ($R^2$): 86% on unseen data.
• Error Metric (MAPE): 28% Mean Absolute Percentage Error.
• Model Validation on Real-World Data (Zillow Test Cases):

Locality	Property Type	Zip Code	Coordinates (Lat, Long)	Sqft	Baths	Beds	Actual Price (Zillow)	Model Prediction	Accuracy
Manhattan	Condo	10021	40.7661, -73.9641	2,200	3	3	$3,850,000	$3,371,963	87.5%
Brooklyn	Townhouse	11211	40.7105, -73.9625	1,150	2	2	$1,250,000	$1,271,603	98.3%
Queens	Multi-family	11103	40.7668, -73.9161	950	1	2	$875,000	$600,413	68.6%

Technical Highlights

• Feature Engineering: Implemented Haversine Distance algorithms to calculate proximity to major NYC landmarks (Times Square & Wall Street), capturing the "Location, Location, Location" factor.
• Data Pipeline: Built a robust ColumnTransformer pipeline for automated encoding of categorical variables and scaling of numerical features.
• Advanced Modeling: Leveraged Random Forest Regressor with Log Transformation on the target variable to handle price skewness and improve convergence.
• Interactive UI: Developed a Streamlit dashboard featuring an integrated Folium map for real-time geographic coordinate selection.

Project Architecture

├── app.py                  # Streamlit Web Application
├── data/
│   ├── raw/                # Original dataset
│   ├── cleaned/            # Cleaned dataset
│   └── processed/          # Featured engineered data ready for modeling
├── models/
│   ├── model.joblib        # Trained Random Forest Model
│   └── preprocessor.joblib # Fitted Scikit-Learn Pipeline
├── notebooks/
│   ├── 01_data_loading_and_overview.ipynb
│   ├── 02_cleaning.ipynb
│   ├── 03_eda.ipynb
│   ├── 04_preprocessing
│   └── 05_modeling
├── requirements.txt        # Reproducibility list
└── README.md

Installation & Usage

1. Clone the repo -> git clone https://github.com/Nourelden85/nyc-housing-prediction.git

2. Install dependencies -> pip install -r requirements.txt

3. Run the App -> streamlit run app.py