Matching Users' Preference under Target Revenue Constraints in Data Recommendation Systems.

Area of Science:

• Information Science
• Computer Science
• Data Science

Background:

• Data recommendation systems aim to match user preferences with available data.
• Balancing user satisfaction with system revenue is a key challenge in recommender systems.

Purpose of the Study:

• To develop an optimization framework for data recommendation strategies.
• To align recommendation mechanisms with user behavior and system revenue goals.

Main Methods:

• Formulating the problem as an optimization task.
• Designing a recommendation mechanism based on relative entropy and utility distribution.
• Utilizing the Message Importance Measure (MIM) for optimal recommendation distribution.

Main Results:

• The optimal recommendation distribution is shown to follow the normalized Message Importance Measure (MIM).
• An 'importance coefficient' parameter allows adjustment for varying system requirements and data distributions.
• The framework demonstrates a method to balance user preference and system revenue.

Conclusions:

• The Message Importance Measure (MIM) provides a rational framework for data recommendation.
• The proposed method offers a flexible approach to data recommendation by adjusting the importance coefficient.
• This work elucidates the physical meaning of MIM in the context of data recommendation.