Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Benchmarking Deep Neural Networks for Modern Recommendation Systems

Abderaouf Bahi, Ibtissem Gasmi

TL;DR

The paper benchmarks seven deep neural architectures (CNN, RNN, GNN, Autoencoder, Transformer, NCF, Siamese) for item-item recommendations across Retail Rocket, Amazon, and Netflix Prize datasets, evaluating both accuracy and diversity. It finds GNNs and Transformers excel in accuracy on e-commerce-like data, while RNNs capture temporal dynamics crucial for Netflix, and Siamese Networks improve diversity, particularly in retail contexts. The study emphasizes the trade-offs between precision and diversity and suggests hybrid approaches that leverage complementary strengths to better satisfy user preferences. Computational demands and data requirements are identified as key practical constraints, framing a path toward efficient, privacy-aware, hybrid recommender systems. Overall, the work advances understanding of how different neural architectures perform under real-world, multi-domain conditions and informs deployment strategies that balance relevance with discovery.

Abstract

This paper examines the deployment of seven different neural network architectures CNN, RNN, GNN, Autoencoder, Transformer, NCF, and Siamese Networks on three distinct datasets: Retail E-commerce, Amazon Products, and Netflix Prize. It evaluates their effectiveness through metrics such as accuracy, recall, F1-score, and diversity in recommendations. The results demonstrate that GNNs are particularly adept at managing complex item relationships in e-commerce environments, whereas RNNs are effective in capturing the temporal dynamics that are essential for platforms such as Netflix.. Siamese Networks are emphasized for their contribution to the diversification of recommendations, particularly in retail settings. Despite their benefits, issues like computational demands, reliance on extensive data, and the challenge of balancing accurate and diverse recommendations are addressed. The study seeks to inform the advancement of recommendation systems by suggesting hybrid methods that merge the strengths of various models to better satisfy user preferences and accommodate the evolving demands of contemporary digital platforms.

Benchmarking Deep Neural Networks for Modern Recommendation Systems

TL;DR

The paper benchmarks seven deep neural architectures (CNN, RNN, GNN, Autoencoder, Transformer, NCF, Siamese) for item-item recommendations across Retail Rocket, Amazon, and Netflix Prize datasets, evaluating both accuracy and diversity. It finds GNNs and Transformers excel in accuracy on e-commerce-like data, while RNNs capture temporal dynamics crucial for Netflix, and Siamese Networks improve diversity, particularly in retail contexts. The study emphasizes the trade-offs between precision and diversity and suggests hybrid approaches that leverage complementary strengths to better satisfy user preferences. Computational demands and data requirements are identified as key practical constraints, framing a path toward efficient, privacy-aware, hybrid recommender systems. Overall, the work advances understanding of how different neural architectures perform under real-world, multi-domain conditions and informs deployment strategies that balance relevance with discovery.

Abstract

This paper examines the deployment of seven different neural network architectures CNN, RNN, GNN, Autoencoder, Transformer, NCF, and Siamese Networks on three distinct datasets: Retail E-commerce, Amazon Products, and Netflix Prize. It evaluates their effectiveness through metrics such as accuracy, recall, F1-score, and diversity in recommendations. The results demonstrate that GNNs are particularly adept at managing complex item relationships in e-commerce environments, whereas RNNs are effective in capturing the temporal dynamics that are essential for platforms such as Netflix.. Siamese Networks are emphasized for their contribution to the diversification of recommendations, particularly in retail settings. Despite their benefits, issues like computational demands, reliance on extensive data, and the challenge of balancing accurate and diverse recommendations are addressed. The study seeks to inform the advancement of recommendation systems by suggesting hybrid methods that merge the strengths of various models to better satisfy user preferences and accommodate the evolving demands of contemporary digital platforms.

Paper Structure

This paper contains 21 sections, 5 equations, 8 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Overview of Recommendation Systems
  4. Neural Network Architectures in Recommender Systems
  5. Accuracy in Recommendation Systems
  6. Diversity in Recommendation Systems
  7. Positioning Our Work in the State of the Art
  8. Method
  9. Data Preprocessing
  10. Model Selection
  11. Model Implementation
  12. Experimentation and Results
  13. Dataset (RQ1)
  14. Evaluation Metrics (RQ1, RQ2)
  15. Justification of Model Parameters (RQ1)
  16. ...and 6 more sections

Figures (8)

  • Figure 1: General workflow of the proposed approach.
  • Figure 2: Metrics visualization on Retail Rocket E-commerce dataset
  • Figure 3: Metrics visualization on Amazon dataset
  • Figure 4: Metrics visualization on Netflix Prize
  • Figure 5: Accuracy and intra-list diversity visualization on top k recommendations
  • ...and 3 more figures