Design And Implementation Of Collaborative Filtering Recommendation Algorithm

Collaborative filtering recommendation technology is one of the most successful and the most widely used personalization recommendation technology which is applied in many fields.CF is predicting the active users' consumption behavior based on similar users' consumption behavior.CF has been widely used in e-commerce,social networks and academic information refer to web 2.0 services.Such as Amazon,Net-flix,e Bay and other web sites have used collaborative filtering algorithms for recommendation.However,the traditional collaborative filtering exist problems,such as less extensibility,sparse datasets and cold-start.With the rapid development of e-commerce,the extensibility problem is particularly prominent,It is difficult to meet the needs of users when the growth of users and items to a certain degree because of the limitation of computing resources and speed.How to solve the recommendation system extensibility problem has become a large challenge.In addition,each user in the recommendation system involved in the project to the proportion of the total number of the project is very small,usually less than 1% or less,thus caused the user data for project evaluation is sparse and serious influence the quality of recommendation.How to solve sparse datasets is also research hotspot in recommendation systems.Aim to solve the less extensibility problem and sparse datasets problem.According to the user clustering and the mixed clustering based on relationship between user and project,we put forward two novel algorithms: mixed collaborative filtering recommendation algorithm based on clustering model and memory,joint clustering collaborative filtering recommendation algorithm based on relationship between user and project.Firstly,according to the less extensibility problem and poor efficiency problem,we proposed hybrid collaborative filtering recommendation algorithm based on clustering model and memory by introducing the clustering algorithm introduced into the traditional collaborative filtering algorithm.The algorithm can improve the efficiency of recommendation and extensibility of system by reducing the time to search the nearest-neighbors customers in the way of offline clustering.In addition,in view of the sparse datasets problem,we propose a joint clustering collaborative filtering recommendation algorithm based on relationship between user and project.The algorithm establish joint clustering model by combine users-projects scoring record,users-users social relations and relationship between projects,use the clustering model realize multidisciplinary comprehensive recommendations.The algorithm significantly reducing the impact of the sparse data and improve the quality of the recommendation of the collaborative filtering recommendation algorithm.