每次使用 mahout 时都有不同的建议
有人可以帮助我吗？
如果我已经向某些用户推荐了一些项目，并且假设该用户对这些推荐的项目不感兴趣，并且他希望看到更多与旧的推荐不同的推荐。
mahout 中是否有任何可用的机制或概念可以帮助我获得新的推荐，而不是已经推荐给用户的旧推荐......
Can anybody help me ?
If i already recommend some items to some user and suppose that user is not interested in those recommended items and he want to see more recommendations which are different from old one's.
Is there any mechanism or concept available in mahout which helps me to get new recommendation's other than old ones which were already recommended to user ......
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
是的。一种方法是将其解释为相似评级。 “不感兴趣”意味着某种程度的低评价。这很简单，而且大部分都有效。但是我不喜欢这种方法。这取决于您的应用程序，但通常“不感兴趣”意味着“不相关”。最好不要将其视为评级。
更好的方法是向
recommend()提供一个Rescorer对象，该对象会过滤掉用户标记为“不感兴趣”的所有项目 ID。Yes. One way is to interpret this as like ratings. "Not interested" means a low rating of some kind. This is easy, and mostly works. However I don't like the approach. It depends on your application, but usually "not interested" means "irrelevant". It's best not to treat that as a rating.
The better way is to supply a
Rescorerobject torecommend()which filters out all item IDs that the user has marked as "not interested".您可以做的一件事（此答案中建议）是随机选择“足够好”的建议子集”，并展示它们。这样每次使用推荐器时，都会返回不同的答案。它并不完美 - 如果没有很多推荐，它们总是会被返回（取决于子集的大小），但它是一个简单的解决方案，比使用 Rescorer 和标记用户不感兴趣更简单。
One thing you can do (which is suggested in this answer) is to randomly pick a subset of recommendations that are "good enough", and present them. That way each time the recommender is used, different answers will be returned. It's not perfect - if there aren't many recommendations, they will always be returned (depending on the size of your subset) but it's a simple solution, more simple than using a Rescorer and marking user disinterest.