Video games recommendation system using collaborative filtering / Tuan Ahmad Wafiq Tuan Mahmud
The video game recommendation system developed in this project addresses the challenge of game discovery due to the vast selection available on platforms such as Steam, Epic Games, and Ubisoft. With over 73,000 games on Steam, gamers struggle to find titles that align with their interests, particula...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/115279/1/115279.pdf https://ir.uitm.edu.my/id/eprint/115279/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The video game recommendation system developed in this project addresses the challenge of game discovery due to the vast selection available on platforms such as Steam, Epic Games, and Ubisoft. With over 73,000 games on Steam, gamers struggle to find titles that align with their interests, particularly since many platforms lack built- in recommender systems and lesser-known games receive minimal exposure. To solve this, the system utilizes collaborative filtering with Non-Negative Matrix Factorization (NMF) to analyze user preferences and provide personalized recommendations. The system was built using a dataset from Kaggle with 200,000 records, including user interactions, playtime, and ratings. Through data preprocessing and filtering, the model ensures accurate recommendations while continuously adapting to evolving user behaviors. Performance evaluation metrics, including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Recall@5, validate the system's effectiveness, demonstrating superior accuracy compared to existing approaches. The results show that NMF-based recommendations significantly improve game relevance and user satisfaction while helping game developers increase the visibility of lesser- known titles. Despite its success, limitations such as the cold-start problem, data sparsity, and computational constraints remain challenges, potentially affecting recommendation accuracy for new users. To enhance the system, future improvements include integrating content-based filtering, expanding datasets, optimizing model training, and leveraging cloud computing for better scalability and real-time recommendations. Additionally, incorporating hybrid recommendation techniques and real-time user feedback could further refine accuracy and relevance, ensuring users receive high-quality suggestions based on their evolving preferences. By improving game discoverability, enhancing user engagement, and providing a scalable solution for other domains requiring personalized recommendations, this project makes a valuable contribution to the field of recommendation systems, benefiting both players and developers in the gaming industry. |
|---|