Abstract | ||
---|---|---|
Current movie recommender system is hard to capture user's preference due to the multidimensional and dynamic characteristics. Aiming at this problem, in this paper, we propose a dynamic hybrid movie recommender framework which models user's preference from four different aspects. The framework is organized according to the classic two-stage information retrieval dichotomy: first, we adopt a suitable recommender algorithm for each aspect respectively for candidate generation, and then a linear combination model is designed to produce the final recommendation list. In order to capture the dynamics of user's preference, We also constructe a feedback learning mechanism which utilize the utility function to compute the best weight vector for each recommender algorithm. Case study on our framework shows that our model can accurately capture user's current interest with acceptable cost. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1007/978-3-319-49178-3_37 | ADVANCES IN SERVICES COMPUTING |
Keywords | Field | DocType |
Hybrid algorithm,Recommender system,Feedback learning,Utility function,Multidimensionality | Recommender system,Linear combination,Hybrid algorithm,Computer science,Weight,Artificial intelligence,Machine learning,Distributed computing | Conference |
Volume | ISSN | Citations |
10065 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 19 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Xiangyong Liu | 1 | 2 | 2.41 |
Guojun Wang | 2 | 437 | 47.52 |
Wenjun Jiang | 3 | 356 | 24.25 |
Yinong Long | 4 | 1 | 0.79 |