- Social Networking Service: an online service, platform, or site that focuses on facilitating the building of social networks or social relations among people who, for example, share interests, activities, backgrounds, or real-life connections.
- Social Media: a group of Internet-based applications that build on the ideological multi faceted and technological foundations of Web 2.0, and that allow the creation and exchange of user-generated content.
- Social Computing: a general term for an area of computer science that is concerned with the intersection of social behavior and computational systems.
- Social Task: a larger view of collaborating work in a social network.
- Social Media Marketing: center on efforts to create content that attracts attention and encourages readers to share it with their social networks.
When we come to the case study of PowerReview, I found that Social Recommendations are very interesting. There are two types of Social Recommendations. The first one is Direct Social Recommendations. Which is simple and directly. And the second one is Derived Social Recommendations, which is more complicated and intelligent. Here I focus on the Derived Social Recommendations, and find some information after class.
Collaborative filtering methods are based on collecting and analyzing a large amount of information on users’ behaviors, activities or preferences and predicting what users will like based on their similarity to other users. There are two types of collaborative filtering, first one is User-based filtering, also named Memory-based filtering, the results of user-based filtering depends on the people who share similar taste with you. If they like the item, then, maybe you prefer it too. For example, if user A like item a, user B like item a, b and c, user C like a and c, then we think user A is similar to user B and C because all of them like item a. Because user B and C like item c too, so the system will then recommend item c to user A. The second one is Item-based filtering, also called Model-based filtering. The results of item-based filtering depend on your previous preferred items. For example, if user A like item a, and item b is similar to a, then, the system will commend item b to user A.
I think the biggest difference between these two filtering methods is that item-based filtering depends on the user’s own taste while user-based filtering depends on others'. In my opinion, the results of item-based filtering maybe more accurate than user-based filtering, because the parallel among items is more stable than users. The chance of different users liking same items is relatively low.
Here we come to Content-based filtering. It’s a method that based on information about and characteristics of the items that are going to be recommended. Whether an item will be recommended to the user depended on the scores calculated according to these preferences and characteristics.
Maybe we could combine the content-based filtering with collaborative filtering in order to get improved recommendations. So, if you are interesting in this aspect, plz leave a comment, so we can have a discussion and exchange ideas.
It's interesting to study the theory of Social Networking while experiencing the relevant famous websites. I will make a list of these websites later.
Welcome to comment on my blog, and we can make progress together.
REFERENCES
- Social Networking Service http://en.wikipedia.org/wiki/Social_Networking
- Social Computing http://en.wikipedia.org/wiki/Social_computing
- Social Media Marketing http://en.wikipedia.org/wiki/Social_media_marketing
- Badrul Sarwar , George Karypis , Joseph Konstan , John Reidl, Item-based collaborative filtering recommendation algorithms, Proceedings of the 10th international conference on World Wide Web, p.285-295, May 01-05, 2001, Hong Kong, Hong Kong
I am very interested in the collaborative filtering methods! As we learnt from lecture 2, we know there are several filtering methods that can recommend people the things they might be interested. Actually, it's difficult to understand the User-based filtering and Item-based filtering before Prof. Chan gave the 4th object to explain the differences between them.
回复删除A good algorithm of recommendation for customers should involve much more complex filtering methods so that it can be optimized and give the most relative and interesting things for users. That can make a better user experience.
It's great that you are interested in this aspect. Prof. Chan's example is very useful, and I found that CUI Helei has written something about it in his blog(http://cuihelei.blogspot.hk/2012/09/the-difference-among-three.html) which is intuitive and helpful.
删除Thank you for your share!
删除Well, you have talked a lot about the filtering methods in social media. I agree most of your points about user-based filtering and item -based filtering, but I think the former one is more progressive and the latter is conservative, so in my opinion, it is really difficult to decide which one is better, maybe depends on which service the social media provides.
回复删除Hi Alfred, your point is very interesting. I hadn't thought in that way. Maybe the item-based filtering is conservative, but the item-based filtering algorithm is more computationally efficient. For more detail, you can have a look at the paper named 《Item-based Collaborative Filtering Recommendation Algorithms》. I wish it will help you.
删除此评论已被作者删除。
回复删除Yeah, I am interested in collaborative filter methods too~ In you example, you said that the results of item-based filtering maybe more accurate than user-based filtering, but I don't think so. User-based filtering is based on the analyzing the similar interest between two users then making the recommendation.It can do more meaningful recommendation.In addition, we can also add some weight, such as trusty weight during the calculation the matrix of user-based filtering
回复删除Your point of view is reasonable, the reason why I think the results of item-based filtering maybe more accurate than user-based filtering is that the parallel among items is more stable than users, also the chance of different users liking same items is relatively low. However, maybe I have jumped to conclusions. It's not that simple.
删除Thank you for your explanation for the filtering methods and your examples. And according to your article, I know the category of the derived Social Recommendations which you said. I think it's interesting opinion and helpful to me to understand the filtering methods in a new way.
回复删除This two kinds of methods of collaborative filtering is interesting. I totally agree with your opinion that the biggest difference between these two methods is that the first one is based on user’s own taste while the other is based on others’. A perfect recommendation mechanism should include all these two methods I think. This collaborative filtering will definitely help us to gain information in a more efficient way.
回复删除Hi, I am very happy that you are agree with my point. It's great. A good filtering method is a powerful tool for social networking analysis.
删除