🚀 Come Al Paso: Restaurant Recommendation System

Come Al Paso is a restaurant recommendation system designed to enhance tourists’ dining experiences by integrating their personal preferences and real-time location. Using a hybrid recommendation approach, the system combines collaborative filtering and content-based techniques to suggest the most suitable restaurants.

The application was developed using Angular 16, Node.js, Express, Firebase, Google API, and IBM Watson's Natural Language Understanding. Its key features include:

• Intuitive Navigation Interface – Allows users to explore features and specify preferences such as restaurant ratings, price range, and cuisine type.
• Personalized Recommendations – Provides a curated list of top-rated restaurants based on Google reviews and sentiment analysis of visitor feedback. The system ensures that recommendations match the tourist’s saved preferences, considering only open restaurants within a defined search radius.
• Detailed Restaurant Information – Displays essential details such as operating hours, contact information, website, specialties, price range, services, and location.
• User Authentication – Supports social media login through Facebook and Gmail.
• Adjustable Search Radius – Allows users to modify the search radius based on their current location.
• Navigation Routes – Provides walking and driving directions from the user’s location to the selected restaurant.

📷 Screenshots

Application with restaurant recommendations and navigation routes Application with advanced filtering options Detailed restaurant information

🖥 Installation and Setup Instructions

Prerequisites

Before setting up the application, ensure you have the following:

• A code editor installed.
• A Firebase account with Authentication and Firestore Database enabled.
• A Google Cloud API Key with the new Places API, Details API and Routes API enabled.
• An IBM Natural Language Understanding API Key.

Installation

Clone the repository and open the project in your preferred code editor.

Install the necessary dependencies:

npm install

Configuration

Server-Side Setup

1. Create Configuration Files:

   In the server folder, create two files:

   • firebase.json
   • .env

2. Configure Firebase Credentials:

   In firebase.json, add your Firebase API credentials:

   {
     "type": "",
     "project_id": "",
     "private_key_id": "",
     "private_key": "",
     "client_email": "",
     "client_id": "",
     "auth_uri": "",
     "token_uri": "",
     "auth_provider_x509_cert_url": "",
     "client_x509_cert_url": "",
     "universe_domain": ""
   }
   

3. Set Up Environment Variables:

   In the .env file, add the API keys provided by Firebase, Google Cloud, and IBM:

   GOOGLE_APPLICATION_CREDENTIALS= PATH_TO_FIREBASE_JSON_FILE
   GOOGLE_API_KEY= YOUR_GOOGLE_API_KEY
   IBM_API_KEY= YOUR_IBM_API_KEY
   
   

Client-Side Setup

1. Configure API Keys and Backend URL:

   In the environment file, add your Google Cloud API Key and Firebase configuration:

   export const environment = {
     production: false,
     GOOGLE_API: '',
     BASE_URL: 'http://localhost:3000', // Update if necessary to match your server configuration
     FIREBASE: {
       apiKey: "",
       authDomain: "",
       projectId: "",
       storageBucket: "",
       messagingSenderId: "",
       appId: ""
     }
   };

Running the Application

Start the backend server:

npm start

Start the frontend server:

cd client
ng serve

Once the servers are running, access the application at:

http://localhost:4200

🖋 Reflection

The primary goal of this restaurant recommendation system was to provide a personalized dining experience tailored to the tourist’s preferences. The application offers an intuitive interface, allowing users to specify their dining criteria, access detailed restaurant information, explore routes, and receive dynamic recommendations.

One of the main challenges was choosing the right recommendation technique. Each approach had limitations, such as cold start issues or item sparsity. To address this, a hybrid algorithm was developed.

The algorithm starts by retrieving the user's location and search radius to identify the top-rated open restaurants on Google. It then ranks these restaurants based on their overall rating, which is determined by analyzing customer reviews through sentiment analysis. If the tourist has preferences, the system filters the restaurants accordingly, prioritizing the three most relevant preferences. Finally, the refined list of recommended restaurants is presented. If no preferences are specified, the system returns the ranked list as is.

The development followed the Scrum framework, allowing for efficient project management and adaptability to tourist needs. Each sprint lasted two weeks, with incremental product releases based on prioritized requirements. Continuous user feedback and testing were conducted at the end of each sprint to enhance the application's functionality.

From a technological standpoint, the system architecture was built using Angular and Express, while the IBM Watson API powered the sentiment analysis module, significantly enhancing the accuracy of recommendations. Additionally, the Google API was integrated to deliver real-time restaurant data, ensuring up-to-date availability and precision. At the conclusion of the development process, a usability test was conducted with 20 participants, yielding a usability score of 81, which indicates a satisfactory user experience. Furthermore, the insights gathered will guide future improvements and refinements to the application.