Research On Recommendation Algorithms Based On Collaborative Filtering In Social Networks

The rapid development of social networks brings serious information overload problems.It is very difficult for users to find the required services or products from a large amount of information.The search engine can help users filter unwanted information,but it need users to input the explicit queried words and it also returns the same recommended results to all users.In order to meet the personalized needs of users and realize the strategy of active recommendation,the recommendation system was born.As the earliest information filtering technology proposed in recommendation system,collaborative filtering has been widely used in the field of e-commerce for it's advantages,such as it does not require domain knowledge of the item and can take advantage of group wisdom.However,the traditional collaborative filtering recommendation algorithm only uses sparse user ratings data for recommendation,resulting in poor recommendation effect.In particular,cold-start users can't obtain recommendations that meet their needs,which seriously restricts the application of collaborative filtering.In order to alleviate the data sparsity problem faced by traditional collaborative filtering,this thesis proposed two novel recommendation algorithms based on collaborative filtering by using the trust relations between users in social networks.The main work of this thesis is as follows:Aiming at the user-cold start problem existing in the traditional collaborative filtering algorithm,a social recommendation algorithm based on collaborative filtering is proposed.It applies the user trust value to replace the position of the user's similarity in the nearest neighbor selection,thereby the problem of poor recommendation results from the inaccurate user's similarity is solved.The algorithm predicts an user's rating to an item by weighting the user's trusting friend's rating,the item's reputation value,and the user's scoring preferences.Finally,simulation comparison experiments on two real datasets show that,compared with other comparison algorithms,this algorithm can provide more accurate recommendations for all users and cold-start users,which proves that this algorithm can effectively mitigate data sparsity and user cold-start problems.Aiming at the poor performance in personalized recommendation of the traditional collaborative filtering recommendation,this thesis further proposes a classification recommendation algorithm based on collaborative filtering,which is on the basis of the social recommendation algorithm.This algorithm is essentially a combination recommendation algorithm which contains trust-based collaborative filtering,user-based collaborative filtering and item-based collaborative filtering.Firstly,we modify the traditional user similarity calculation formula to obtain a similarity calculation formula with higher accuracy.Then,according to the characteristics of the target user and the target item,we select the most suitable factors from the similar users,trusting users,similar items,item reputation value,user's scoring preferences to predict an user's rating to an item.Finally,we have done the simulation comparison experiment on the algorithm.Results show that,compared with the popular matrix factorization-based recommendation algorithm and trustbased recommendation algorithm,this algorithm can provide more accurate recommendation results and meet user's personalized needs.