Advantages Of Collaborative Filtering

ABSTRACT

Recommendation systems has gain utmost importance in the field of research since the emergence of the first paper on collaborative filtering in the mid-1990s.The most widely used approach for recommendation system, known as collaborative filtering is based on past user behavior such as product rating or previous transactions. Collaborative filtering was first defined by the tapestry developers, the inventor of first recommendation system. Collaborative Filtering aims at analyzing the interdependencies between products and the relation among users in order to recommend items to users. A major advantage of collaborative filtering algorithm is that it does not require the collection of large amount of external data that is not easily

To make the recommendations more personalized the web giants uses the social [15] data such as the friendship relations, strength of links between two or more users. It is beneficial for both user and the providers. The users need not waste time in searching for desired items and as a result increases the sales or hits for the providers.

Recommender systems or recommendation systems forms a class of data and information filtering system that aids in predicting the taste and preferences of a user. It has gain importance in recent years and is being actively used in various applications. The most common ones are music, research papers, books, movies, jokes, social tags, and products

1.1 HISTORY OF RECOMMENDER SYSTEM

The availability of humongous amount of data on internet let to the necessity if a recommender system Grundy [17], an automated computerized librarian was a first step toward recommender system It was a simple traditional grouping of users into “stereotypes” based on an interview and using this information for recommending

It depends only upon the average preference difference between every pair of items, which can be pre- computed. Moreover, we can easily update its underlying data structure without affecting the whole system i.e. when a preference for a particular item is changed, its corresponding values can be updated easily.

One of the main drawbacks of slope one algorithm is that it is memory intensive. The memory required to store all precomputed differences between item pairs preference value is equal to the square the number of items. Twice as many items means four times the memory!

User-Similarity

As the name suggest, here similarity between two user vectors is computed using various similarity measures and user’s preferences recommendations are made according to those measures. User-

based recommenders are the traditional style of recommender systems. They are mainly used for small data sets (roughly, less than 10 million ratings). The basic idea behind this approach is that in order to compute recommendations for any of the particular user, we look for other users having a similar taste and preferences and we use their liked items to provide the