CFDIL: a context-aware feature deep interaction learning for app recommendation.

Area of Science:

• Computer Science
• Human-Computer Interaction

Background:

• Mobile applications (apps) have proliferated, leading to user app overload.
• Traditional app recommendation methods struggle with sparse data and overlooked contextual information.

Purpose of the Study:

• To propose a novel context-aware feature deep interaction learning (CFDIL) method for enhanced app recommendation.
• To address data sparsity and incorporate contextual information in user preference modeling.

Main Methods:

• CFDIL constructs user and app feature portraits to incorporate contextual features.
• Dense feature portraits and tensor operations on label sets mitigate data sparsity.
• A deep network structure is trained using contextual and attribute information for accurate recommendations.

Main Results:

• CFDIL effectively models user preferences by integrating contextual features.
• The method demonstrates superior performance compared to benchmark approaches on three real datasets.
• Data sparsity challenges are overcome through dense feature representations and tensor operations.

Conclusions:

• CFDIL offers a significant advancement in app recommendation systems.
• Incorporating context-aware features and addressing data sparsity are crucial for improving recommendation accuracy.
• The proposed deep learning approach provides a robust solution for the app overload problem.