🏠 Predicting Apartment Prices in Buenos Aires

📘 Project Overview

This project aims to predict real estate prices in Buenos Aires using a dataset of apartment listings. It focuses on using multiple features — including size, location, and neighborhood — to train a regression model that can estimate property prices with accuracy.
The project demonstrates data cleaning, feature engineering, and model building in Python.

---

🧰 Tech Stack

• Python
• Pandas, NumPy – Data manipulation
• Matplotlib, Seaborn – Visualization
• Scikit-learn – Model training and evaluation
• Category Encoders – Encoding categorical variables
• Jupyter Widgets – Interactive exploration

---

🔍 Key Steps

1. Data Preparation

• Loaded and cleaned multiple CSV files.
• Filtered data to include only apartments in Capital Federal under $400,000.
• Removed outliers in surface_covered_in_m2.
• Extracted latitude, longitude, and neighborhood information.
• Dropped columns with >50% missing values.

2. Feature Engineering

• Handled categorical variables with low or high cardinality.
• Prevented data leakage by dropping features derived from the target.
• Removed strongly correlated features (correlation > 0.9).

3. Model Development

• Defined a feature matrix (X) and target variable (y).
• Split dataset into training and validation sets.
• Trained multiple regression models to predict prices.
• Evaluated models using metrics such as R², MAE, and RMSE.

4. Results & Insights

• The model successfully learned relationships between location, size, and neighborhood and apartment prices.
• Demonstrated the impact of feature selection and data cleaning on model performance.

---

📈 Example Features Used

Feature	Description
surface_covered_in_m2	Covered area in square meters
rooms	Number of rooms
neighborhood	Extracted from parent names
latitude / longitude	Property coordinates
property_type	Type of property (e.g., apartment)

---

🚀 How to Run

1. Clone the repository or download the notebook.
2. Install required packages:

   pip install pandas numpy matplotlib seaborn scikit-learn category_encoders

3. Open the notebook:

   jupyter notebook predicting_price_with_multiple_fatures.ipynb

4. Run all cells sequentially to reproduce results.

---

📊 Future Improvements

• Experiment with XGBoost or Random Forest Regressors for better performance.
• Integrate geospatial visualization (e.g., Folium).
• Deploy model using Streamlit for interactive prediction.

---

✨ Author

Pushkin Kumar
Data Analyst | Data Engineer
Passionate about building modern data pipelines and predictive models.