Dynamic User Preferences Optimization in Time-Aware Recommendation Systems
Ebrahimi, Ainaz (2024)
Ebrahimi, Ainaz
2024
Master's Programme in Computing Sciences
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2024-10-15
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202410109215
https://urn.fi/URN:NBN:fi:tuni-202410109215
Tiivistelmä
The integration of recommendation technology within Internet of Things (IoT) environments has been significantly accelerated by advancements in network technology. These recommendation systems, which predict user preferences to deliver personalized suggestions, play a critical role in mitigating information overload by presenting the most relevant products or services to users. Traditional recommendation algorithms—including collaborative filtering, content-based filtering, and knowledge-based approaches— have proven effective but face challenges related to data sparsity, scalability, and the evolving nature of user preferences.
This research addresses these limitations by enhancing Time Correlation Coefficient (TCC) models to better capture temporal dynamics. Unlike existing models that rely on a single attenuation coefficient, this study introduces customized coefficients tailored to each user, allowing the system to more precisely capture individual preference shifts over time. Furthermore, by integrating the TCC with item similarity scores, the approach aims to enhance both the relevance and accuracy of the recommendations. Four innovative methodologies are proposed to refine the TCC framework: dynamic attenuation coefficient exploration via Time-based Decay, Cuckoo Search Optimization, Decay Model Selection, and Content-based Similarity integration with TCC through item similarity scores.
The dissertation demonstrates substantial progress in temporal collaborative filtering through detailed theoretical analysis and practical validation using three real-world datasets: two from Amazon and one from MovieLens. The results show that the proposed approaches significantly improve the accuracy of the recommendation system compared to current models. By incorporating time-aware adjustments, the study highlights measurable enhancements in recommendation accuracy, underscoring the effectiveness of dynamic, time-based approaches in personalizing recommendation systems. These findings have important implications for future research and practical implementations in personalized recommendation services.
This research addresses these limitations by enhancing Time Correlation Coefficient (TCC) models to better capture temporal dynamics. Unlike existing models that rely on a single attenuation coefficient, this study introduces customized coefficients tailored to each user, allowing the system to more precisely capture individual preference shifts over time. Furthermore, by integrating the TCC with item similarity scores, the approach aims to enhance both the relevance and accuracy of the recommendations. Four innovative methodologies are proposed to refine the TCC framework: dynamic attenuation coefficient exploration via Time-based Decay, Cuckoo Search Optimization, Decay Model Selection, and Content-based Similarity integration with TCC through item similarity scores.
The dissertation demonstrates substantial progress in temporal collaborative filtering through detailed theoretical analysis and practical validation using three real-world datasets: two from Amazon and one from MovieLens. The results show that the proposed approaches significantly improve the accuracy of the recommendation system compared to current models. By incorporating time-aware adjustments, the study highlights measurable enhancements in recommendation accuracy, underscoring the effectiveness of dynamic, time-based approaches in personalizing recommendation systems. These findings have important implications for future research and practical implementations in personalized recommendation services.