Genre-aware user profiling using duration count matrices: A novel approach to enhancing content recommendation...

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Recommender systems are crucial for user experience and content discovery.
• Conventional methods often fail to adapt to evolving user preferences, leading to irrelevant recommendations.
• Capturing long-term user behavior is essential for effective personalization.

Purpose of the Study:

• To develop an innovative method for personalized recommendations using watch-time duration.
• To address the limitations of existing recommender systems in capturing dynamic user interests.
• To enhance content discovery through more accurate and adaptive recommendations.

Main Methods:

• Introduction of the Duration Count Matrix (DCM) technique.
• DCM comprises User Profiling (DCM-UP) for dynamic profile construction and User Similarity (DCM-US) for collaborative filtering.
• Utilizes matrix-based representations and dynamic updates to reflect changing user preferences.

Main Results:

• The DCM approach demonstrated significant outperformance against state-of-the-art methods.
• Evaluated on a real-world dataset from JAWWY, showing improvements in precision, recall, F1-score, and accuracy.
• The technique effectively captures and predicts long-term user behavior for superior personalization.

Conclusions:

• The proposed DCM technique offers a superior method for personalized recommendations.
• Leveraging watch-time duration effectively models long-term user engagement.
• This approach leads to more accurate, adaptive, and relevant content discovery.