Real Estate Listing Recommender

This project showcases a Real Estate Listing Recommender application using a content-based filtering approach with a sentence transformer model. The application allows users to input queries and receive relevant real estate listings based on their descriptions and other textual data.

Demo

Features

• Content-Based Filtering: Recommends listings based on the semantic similarity of the user's query and listing descriptions and past user reviews, using a sentence transformer model.
• Interactive Frontend: A React frontend with a clean, professional look and feel, showcasing the recommendations in a user-friendly interface.
• Backend API: A Flask backend serving the recommendation results, optimized for performance using FAISS for similarity search.
• Deployment: The application is set up for local testing and is demonstrated via a hosted GIF in the README.

Setup and Installation

Prerequisites

• Node.js and npm for frontend.
• Python for backend.
• Git for version control.

Backend Setup (Flask)

1. Clone the repository:

   git clone https://github.com/VarunRavi95/RecSys
   cd RecSys
   

2. Install dependencies

   pip install -r requirements.txt
   

3. Start the Flask server:

   python app.py
   

Frontend Setup (React)

1. Install dependencies:

   npm install
   

2. Navigate to the app folder:

   cd my-react-app
   

3. Start the React application:

   npm start
   

Key Components:

1. Backend:

   • Flask API: Handles requests and processes recommendations using FAISS for efficient similarity search. • Sentence Transformer Model: Utilizes the all-mpnet-base-v2 model to embed listing descriptions and user queries.

2. Frontend:

   • React Application: Provides a user interface for inputting queries and displaying recommendations. • Styling: CSS for a modern, professional look with features like rounded edges and card sliders for a better UX.

Development Process

1. Data Preprocessing: Cleaned and merged relevant textual data from the Airbnb dataset for the Stockholm region.
2. Model Training and Embedding: Created embeddings using a powerful sentence transformer and indexed them with FAISS for fast retrieval.
3. Building the Frontend: Designed a responsive UI with React and CSS to display recommendations.

Future Enhancements:

•	Additional Filters: Implement user filters for more refined recommendations.
•	Expanded Dataset: Integrate more diverse real estate datasets for broader recommendations.
•	Scalable Deployment: Move to cloud-based deployment solutions for better scalability and availability.