Movie Recommendations Based On Collaborative Filtering Algorithm

In recent years,big data has become a major challenge facing the development of the Internet,but also a great opportunity.Today,the world of information explosion in the state,resulting in a large amount of information,poor information quality,information value and low.Information overload problem caused when users are looking for the individual needs of severely low efficiency,for which scientists have proposed a personalized recommendation algorithm.It is based on the user's information needs,interests,characteristics,buying behavior,will direct interested users to information,products and other information to the user's personalized recommendation system.In order to improve the recommendation accuracy personalized recommendation system,researchers have proposed a variety of recommendation algorithm,in which collaborative filtering technology is personalized recommendation system is the most successful and most widely used technique,in theory and in practice have made rapid development of.It is based on the user's selection history information and the similarity relationship,and user interests gather the same information to other users of the evaluation,a recommendation.However,collaborative filtering algorithm faces difficult to measure the similarity of problems affecting the quality of the recommendation system.Similarity measure for the problem of collaborative filtering algorithm corresponding improvements The main work is as follows: user-based collaborative filtering algorithm for the proposed introduction of a factor in the calculation of the user similarity P,P value equal to two user ratings the intersection of two of the project over the project and the user rates set,it is clear that the range of coefficients between 0-1.This will ensure that only participated in the most user ratings,and ratings users about the same project,it will most likely be similar to the user.Exclude the user who only participated in a handful of several ratings and ratings similar.By performing detection based Movie Lens dataset movie rating data,experimental results show that the recommendation did improve accuracy;collaborative filtering algorithm based on the category for proposed similarity calculation formula for a correction,the contribution to the similarity of goods on the basis of active users should not be less than the principle of non-active users,increase the number of user activity on the inverse of this parameter,to punish active users.But for those who are too active users,in order to avoid too dense similarity matrix,generally can not be incorporated into the similarity calculation.By performing detection based MovieLens dataset movie rating data,experimental results show that really improved the accuracy of recommendation.