ABSTRACT

A recommender system is a subclass of information filtering system that is used to predict the value or priority of an item, for example in shopping online. One of the famous classes to recommend an item or value is the collaborative filtering, which presents the product according to the interest/preferences of the active user. Unfortunately, this technique suffers from many problems like sparsity and scalability for huge databases, which contains huge number of users and items. To design an
efficient recommender systems the techniques of union and intersection are most widely used.

This work presents an analysis about collaborative filtering techniques. The techniques used are union, -intersection and neighborhood. For the implementation a local database of movies is created. After the implementation these three techniques, the neighborhood technique provides greater accuracy when compared with union and intersection. The collaborative filtering through neighborhood method also recommends all of the unwatched movies. However in terms of diversity of choice the union based technique provided a better user experience through recommendation of movies from different categories. In terms of popularity, the neighborhood collaborative filtering also includes the more number of less popular movies in the recommendations when compared with the union and intersection techniques.