London-Real-Estate-Investment-Estimator

An End-to-End Machine Learning tool that identifies undervalued properties in London by analyzing Geospatial, Textual, and Physical data.

🚀 The Business Problem

Real Estate investors in London face a massive data problem: Price does not equal Value.

A property listed for $100/night might be a steal, while one at $150 might be overpriced junk.

Traditional analysis looks at simple averages, missing the nuance of “Vibe” (guest reviews) and precise location dynamics.

The Solution: I built an End-to-End Machine Learning Engine that:

1. Estimates Fair Value for any property based on 150+ features.
2. Identifies Arbitrage Opportunities: Properties where Actual Price < Predicted Price (Undervalued Gems).
3. Quantifies “Vibe”: Uses NLP to turn guest sentiment into hard data.

---

📊 Key Results & Insights

• Model Performance: Achieved an R² of 0.82 and MAE of $28, meaning the model explains 82% of the market variance.
• The “Vibe” Reality: Bedrooms and Privacy drive price 3x more than positive sentiment. Guests compliment the host, but they pay for space.
• Location Decay: Properties within 5km of the center command a 40% premium, following a strict exponential decay curve.

1. What actually drives price?

2. The Market Map

3. Streamlit Deployment

Streamlit Deployment

---

🛠️ Tech Stack & Methodology

This project follows a 4-layer Data Science architecture:

Layer	Technology	Description
1. Engineering	Pandas, NumPy	Cleaned 96,000 raw rows, parsed complex text (Amenities), and imputed missing values.
2. Geospatial	Folium, Haversine	engineered Distance_to_Center feature and built interactive cluster maps.
3. NLP	NLTK (VADER)	Analyzed 50,000+ guest reviews to generate a Sentiment_Score (-1 to +1) for every listing.
4. Machine Learning	XGBoost, Scikit-Learn	Trained a Gradient Boosting Regressor to predict price with 82% accuracy.
5. Deployment	Streamlit	Built a user-facing app for investors to test scenarios (e.g., “What if I add a pool?”).

---

📂 Repository Structure

├── app.py                   # The Streamlit Application (Deployment)
├── notebook_analysis.ipynb  # The Core Analysis (Cleaning, EDA, ML Training)
├── xgb_model.json           # Trained XGBoost Model
├── model_columns.joblib     # Saved feature names for consistency
├── app_data.csv             # Lightweight data for the map visualization
├── requirements.txt         # List of libraries
├── README.md                # Project Documentation
└── Visuals/                  # Screenshots for README

Author

Ahmed A. Elatwy

• LinkedIn
• Portfolio

Built as a Capstone Project demonstrating Full-Stack Data Science capabilities.