Context-based Knowledge Recommendation: A 3-D Collaborative Filtering Approach
Liang Kaichun, Shuqin Cai, Qiankun Zhao · 2007
We propose a novel and enhanced knowledge recommendation approach using 3D collaborative filtering. Our approach has the following advantages: (1) rather than only use the user-item matrix, the context of filtering is modeled in the third dimension, which makes the recommendation more accurate; (2) the sparseness of the user-item matrix can be partially solved by propagating ratings among the rating matrix with respect to users' backgrounds; (3) the relations between different contexts are embedded in the recommendation as well. Experiments have been done with real data collected in an enterprise knowledge-base for ill-structured problem solving. The results show that our 3D collaborative filtering approach can improve the existing approaches in terms of both the quality of knowledge recommendation and robustness.