Real_Estate_project-

Oman Real Estate Price Analysis:

Project Objective:

This project simulates a real-world data science task in the real estate domain. The key goals are to:

• Scrape property listings data from two major websites in Oman.
• Clean and integrate the datasets into a consistent format.
• Engineer useful features related to pricing and location.
• Frame and solve a predictive modeling challenge: predicting property prices.
• Deliver a complete GitHub repository containing code, data, and documentation.

Websites Used:

The data was collected from the following real estate platforms in Oman:

1. Dubizzle Oman - Properties for Sale
2. OpenSooq Oman - Property Listings

Only properties for sale were targeted (not rentals).

Steps Taken: Data Collection:

• Used requests and BeautifulSoup to scrape listing pages and individual property details.
• Targeted key fields:
• Title
• Location
• Price
• Size (m²)
• Number of rooms(if available)
• Listing type

Data Cleaning:

• Removed listings with missing or invalid values (e.g., N/A).
• Converted text fields like price and size to numeric.
• Removed top 1% outliers in price and size to reduce skew.
• Merged the datasets from both websites using common columns (location, price, size).

Feature Engineering:

• Created price_per_m2: price divided by size.

• Created log_price: log-transformed price to reduce skewness.

• Encoded location using:

  • Frequency encoding: location_freq
  • One-hot encoding for modeling: location_Muscat, location_Sohar, etc.

  Modeling Approach:

The challenge: Predict property sale price based on available features.

• Model Used: RandomForestRegressor (via scikit-learn)
• Target Variable: log_price
• Features Used:
  • size, price_per_m2, location_freq, and one-hot encoded location variables
• Train-Test Split: 80/20
• Evaluation Metrics:
  • Mean Squared Error (MSE)
  • R² Score