사용자 동적 활동 시간을 활용한 추천시스템에서의 RWR 알고리즘 개선
- Title
- 사용자 동적 활동 시간을 활용한 추천시스템에서의 RWR 알고리즘 개선
- Other Titles
- A Random walk with restart algorithm using user dynamic action time for recommendation systems
- Author
- 김민규
- Alternative Author(s)
- MinGyu Kim
- Advisor(s)
- 이기천
- Issue Date
- 2018-02
- Publisher
- 한양대학교
- Degree
- Master
- Abstract
- A Recommendation system that recommends an appropriate item by predicting a user's preference has been studied for a long time. The existing recommendation system suggests a recommendation algorithm using explicit data that gives feedback on items directly by the user. However, collecting all explicit evaluations of actual data is a real challenge. To overcome these drawbacks, studies using implicit data using metadata of explicit data are actively being conducted. Therefore, we propose a method to improve the performance of recommendation algorithm by using implicit data using time information. In this paper, we apply the time decay function to each user differently to grasp the item preference according to the time of each user. That is, the algorithm uses a method of reducing the weight of the old item among the items selected by the user and increasing the weight of the recently selected item. We also used the Random walk with Restart (RWR) model, which is easy to include implicit data, is robust against data sparsity, and is a graph-based model for personal recommendations. In order to apply the time decay function to both situations where the data is sparse or dense because the data may be scarce and the graph may not be formed, we propose a model that combines the RWR model and the Time decay function. We used three data sets for the algorithm test, and compared with the existing RWR algorithm, we could confirm the higher performance.
- URI
- https://repository.hanyang.ac.kr/handle/20.500.11754/69000http://hanyang.dcollection.net/common/orgView/200000432581
- Appears in Collections:
- GRADUATE SCHOOL[S](대학원) > INDUSTRIAL ENGINEERING(산업공학과) > Theses (Master)
- Files in This Item:
There are no files associated with this item.
- Export
- RIS (EndNote)
- XLS (Excel)
- XML