A collaborative filtering method for interactive platforms
A collaborative filtering method for interactive platforms
复制标题
DOI：
10.1109/icdim.2017.8244679
复制
发表时间：
2017-06
期刊：
影响因子：
--
通讯作者：
Yan Zhou;Taketoshi Ushiama
中科院分区：
文献类型：
--
作者：
Yan Zhou;Taketoshi Ushiama
诸如Last.fm和Steam等互动平台目前在电子商务中发挥着越来越重要的作用。交互式平台中最重要的特征是流数据，其包含关于用户在任何时间的兴趣的大量信息。然而，以前的推荐系统已经无法处理流数据。因此，我们提出了一个协同过滤方法，使用滑动窗口技术。此外，我们发现，滑动只在交互时间的结果在一个更好的性能。此外，我们提出了一种称为等比填充的方法来处理次优的流数据和其他优化策略。最后，我们使用流数据集评估了我们的方法。结果表明，我们的方法比其他传统的方法性能更好。
Interactive platforms such as Last.fm and Steam are currently playing an increasingly important role in ecommerce. The most important feature in an interactive platform is streaming data, which contain an enormous amount of information regarding a user's interests at any time. However, previous recommender systems have been unable to deal with streaming data well. Therefore, we propose a collaborative filtering approach that uses the sliding window technique. Furthermore, we found that sliding only on interaction time results in a better performance. Moreover, we propose a method called equal ratio filling to handle suboptimal streaming data and other optimization strategies. Finally, we evaluated our approach using the stream dataset. As the results indicate, our approach performs better than other conventional approaches.