GitHub Description

A hybrid recommendation system combining High Utility Itemset Mining (EIHI) and advanced recommendation models (CB, CF, GNN) for optimized data processing, improved accuracy, and personalized user experiences.

Hybrid Recommendation System

This project presents the design and implementation of a hybrid recommendation system that combines various approaches to deliver accurate and personalized recommendations.

Overview

The system integrates the following components:
1. High Utility Itemset Mining (EIHI):
- Utilizes the EIHI model to optimize the mining process by reducing the search space and improving data processing efficiency.

1. Combination of CB, CF, and HUI Models:

   • Employs algorithms such as KNN, User-Based, Item-Based, and SVD to enhance recommendation quality and accuracy.
     

2. Graph Neural Networks (GNN):

   • Leverages complex graph features and real-world product values to further optimize recommendations and enhance user experiences.
     

Features

• Data Optimization: Efficient processing through High Utility Itemset Mining.
  
• Accuracy: Enhanced precision with CB, CF, and collaborative HUI methods.
  
• Personalization: Flexible and user-focused recommendations.
  
• Modern Techniques: Integration of GNN to utilize graph-based insights.
  

Benefits

• Improved system accuracy.
  
• Enhanced flexibility to meet diverse user needs.
  
• Scalable and robust architecture for real-world applications.

Experimental Results

The hybrid recommendation system was evaluated on accuracy, revenue, and the impact of integrating High Utility Itemset Mining (HUI).

• Accuracy:

  • SVD achieved the best accuracy with the lowest RMSE (0.9125), followed by User-Based and Item-Based models.
    
  • Content KNN showed the lowest accuracy, limited by its reliance on textual descriptions.
    

• Revenue:

  • HUI integration significantly boosted revenue across all models.
    
  • Item-Based + HUI recorded the highest revenue (41.35 GBP), with an increase of 21.29 GBP.
    

• GNN Performance:

  • GNN + HUI reduced cross-entropy error but showed a slight trade-off in accuracy, highlighting optimization opportunities.
    

HUI integration demonstrated clear improvements in both accuracy and revenue, validating its effectiveness in enhancing recommendation systems.

Technologies Used

• High Utility Itemset Mining (EIHI)
  
• Content-Based (CB) and Collaborative Filtering (CF)
  
• Graph Neural Networks (GNN)
  

Contributing : @ngoctrang315

Let me know if you need help with customization or further instructions!