Improve And Application Research Of Collaborative Filtering Algorithms

With the development of Internet technology,explosive information appears in people's daily life..Exponential growth of data resources makes accurate access to valuable information more urgent.Personalized recommendation technology comes into ours field of view.By collecting the historical behavior of users on the platform,this technology analyzes the internal reason that dominates the behavior,that is possible preferences of the user.Based on this,information service and decision support are actively provided to users with recommendation algorithm.Collaborative filtering algorithm is widely used in the field of personalized recommendation,which is the most widely explored.In this paper,an improved collaborative filtering recommendation algorithm is proposed to solve the problems of scalability and timeliness of recommendation in traditional algorithms,and it is successfully applied to the service platform of small and medium-sized enterprises.The main research contents of this paper are as follows:(1)By studying the traditional collaborative filtering recommendation algorithm based on user,project and model,this paper discusses the similarity measurement method and the criterion of recommendation algorithm,and expounds the problems faced by the algorithm and the urgency of improvement.(2)An improved collaborative filtering recommendation algorithm based on partition algorithm is proposed.Firstly,the partition algorithm is used to divide the database into several disjoint sub-databases,and to find local frequent items.Then we use Pearson's correlation measurement method correlation measure to find out the similar neighbors of top-n.The query then combines the frequent items that contain j items to form a U class.Finally,the first n items in the U class are recommended by the recommended users in the prediction band to evaluate the U class.(3)Experiments on real data sets show that the average absolute error and mean square root error of the algorithm are lower than those of the traditional algorithm,and the coverage of the algorithm is higher than that of the traditional algorithm at the same time.It is proved that the recommendation accuracy and coverage ability of the improved algorithm are better than those of the traditional algorithm.(4)AHP is used to solve the challenges faced by the improved algorithm in practical application.The algorithm is applied to the service platform of small and medium-sized enterprises.In view of the problems such as the platform has not been popularized,the number of users and thedata are not perfect,the user model is established with AHP.On the premise of establishing the hierarchical structure model,the judgment matrix is constructed according to the scale of 1 ~ 9,and the weight vector is obtained by checking the consistency of the judgment matrix.The weight matrix is combined to get the decision weight,and based on the model,the top users in the model are rewarded.Experiments show that the increment of platform users increases and the period of user project decreases after the algorithm is used,which proves that the algorithm has good practical significance.The results show that the proposed algorithm not only has good scalability,but also shortens the recommendation time.It is suitable for the service platform of small and medium-sized enterprises,and promotes the development of public service and entrepreneurship.