Recommender Systems

 Recommender systems

Individual consumers' interests and preferences are elicited by recommendation systems, which then make recommendations based on their findings. They have the potential to help consumers make better decisions when searching for and purchasing products online.They have the potential to aid and improve the quality of consumer decisions made while searching for and purchasing products online.  the most common place I've seen recommender systems are websites. for example amazon. 

a recommender System is a platform used as a subclass of information filtering system that seeks to predict the rating or preferences a user would give to an item. 

for example in amazon by clustering my typical choices based on my buying habits. It can be seen that I have very similar responses and therefore very similar interest and behaviours in these products. Therefore my amazon has decided to recommend me these due to me searching for these regularly.  I search these because I am a hairdresser part time.

Most recommender systems takes either of two basic approaches:

-collaborative filtering

-content-based filtering

Collaborative filtering

Based on a model of prior user behaviour, collaborative filtering generates a recommendation. 

The model can be built entirely out of

 -the behaviour of a single user, or more effectively, the behaviour of other users who share similar characteristics. 

Collaborative filtering uses group knowledge to form a recommendation based on similar users when it takes other users' behaviour into account.

Problems of Collaborative filtering

•Cold Start: There needs to be enough other users already in the system to find a match.

 

•Sparsity: If there are many items to be recommended, even if there are many users, the user/ratings matrix is sparse, and it is hard to find users that have rated the same items.

 

•First Rater: Cannot recommend an item that has not been previously rated.

–New items

–Esoteric items

 

•Popularity Bias: Cannot recommend items to someone with unique tastes. 

– Tends to recommend popular items.

 

Content - based filtering

Uses a machine learning algorithm to infer a profile of the users preferences from examples based on a featureural description of content rather than other users' opinions.

Advantages

- No need for data on other users

- no cold start or sparsity problems

- Able to recommend new unpopular items

-Can provide explanations of Recommended items by listing content - features that caused an item to be recommended.

Disadvantages 

- Requires content that can be encoded as meaningful features. 

-Users’ tastes must be represented as a leasable function of these content features. 

-Unable to exploit quality judgments of other users.(Unless these are somehow included in the content features

Hybrid approach= combines collaborative and content- based filtering therefore increasing the complexity of the recommender system.