Post 2 - Social Recommender System

A new term to me - Social Recommender System, however, in fact is what we had been having , using without your recognition for a long time.


So what is it?
According to wiki-pedia , it is evolved from information filtering system, to predict the 'rating' or 'preference' that a user would give to an item in a systemic way. By performing analysis (like Collaborative filtering / Content-based filtering ) , a generalized set of recommendations will be resulted.

This is widely adopts is many social platforms, which contribute the functions like 'number of reviews' , 'top rated', 'Most view', 'Most response',etc.

Type of Social Recommendations:
Direct Social Recommendations: emphasize direct relationships, like a direct comment, direct reviews. 

Derived Social Recommendations: relies on derived a complex software analysis to generate the recommendations, like most read / most recommended, mutual friends, etc. 

Approach to Social Recommendations:    

There are 4 main approaches to recommendations: 

1. Personalized recommendation - recommend things based on the individual's past behavior

Personalized Recommendations - Google Buzz

 
2. Social recommendation - recommend things based on the past behavior of similar uses 
3. Item recommendation - recommend things based on the item itself  
4. A combination of the three approaches above

Item Recommendations - Amazon Personalized Recommendations

Why will need a recommender system?
Information is overloaded with the grow of number of users. When a social applications grow with larger number of users, information / interactions among users will grow significantly. This created the problem of information overload and may even discourage user experiences. 
This make finding relevant content for a users become a important and worth-investing for improvement item in social computing and research area.
Recommeder system provide a key solution for solving this problem.



Recommender system also to many business opportunity. Like sales prediction, identify of potential customer help to identify their potential via different algorithms.





Problem face by recommender system:
1. Lack of Data

- To provide a accurate result, a large amount of data is required. Take example of items recommendations of Netflix, it takes a large user base and thus created a very accurate recommendations. The more user data you get, the easier you can get a accurate recommendations.

2. Changing Data

Data is always changing and Trend is always changing too.

Take an example of a fashion shop website called StyleHop, a resource and community for fashion enthusiasts. The items is always changing and the fashion is always changing.

Simply analyzing the users pass behavior will not work. Also items recommendations won't work because each item (like a clothing) have too many attribute and characteristics and also they too many items!

 

3. Changing User Preferences

Social recommendation is trying to predict which users needs and give a corresponding recommendations or suggestion. However, it is also impossbible to have a accurate prediction on users preferences. Like one day I will be browsing Amazon for new books for myself, but the next day I'll be on Amazon searching for a birthday present for my sister.

  

4. Unpredictable Items

Some items are hard to predict in nature, or have no a objective rule for rating. One typical examples is your favorite song.

References:
[1] Richard MacManus / January 26, 2009 / A Guide to Recommender Systems

[2] Ido Guy, David Carmel / March 2011 / Social Recommender Systems

[3] Richard MacManus / January 28, 2009 / 5 Problems of Recommender Systems