📊 Predicting Apartment Prices in Mexico City

🏙️ Project Overview

This project predicts the prices of apartments in Mexico City (Distrito Federal) using multiple datasets. The data undergoes extensive cleaning, preparation, and modeling to build a regression model that forecasts apartment prices based on size and location.

📂 Data

• Five CSV files containing real estate listings.
• Data includes apartment size, price, location (borough), and more.

🛠️ Tools & Libraries

• Python
• Libraries: -Pandas, NumPy, Seaborn, Matplotlib, Scikit-Learn, Category Encoders

📋 Data Processing Workflow

1. Import & Inspection
   • Read CSV files and inspect structure.
   • Identify columns with missing values and drop those with ≥ 50% null values.
2. Filtering
   • Focus on apartments (property_type).
   • Filter properties located in Mexico City (Distrito Federal).
3. Feature Engineering
   • Extract latitude (lat) and longitude (lon) from lat-lon column.
   • Create a new borough column from location hierarchy.
4. Handling Outliers
   • Apartments priced > $150k are excluded.
   • Filter apartments within the 10th-90th percentile for area (sq meters).
5. Categorical Columns
   • Drop low/high cardinality columns (operation, currency, etc.).
6. Target Leakage
   • Remove columns like price, price_aprox_local_currency, and price_per_m2 to avoid data leakage.
7. Final Dataset
   • Concatenate cleaned data from all CSV files.

📊 Exploratory Data Analysis

• Histograms visualize price distribution (right-skewed).
• Scatter plots show price vs. apartment area.

📈 Model Building

• Baseline Model: Predicts the average apartment price for all observations.
• Pipeline: Combines imputation, encoding, and regression.
• Linear Regression used as the primary model.

🚀 Results

• The model’s performance is evaluated using Mean Absolute Error (MAE).
• Results indicate the model can effectively estimate prices for affordable apartments in Mexico City.