Research On Spam Information Filtering And Recommender System Of Social Network

With the rapid development of Internet technology,social networks have become an important part of people's lives and have brought great convenience to people.Social networks also attract merchants' campaigns with their large user base,agile and open information dissemination.However,the existing social network platform lacks a formal and friendly product recommendation mechanism and a spam screening mechanism,which leads to social networks filled with various kinds of advertisement information,among which there are many fake and shoddy goods.These problems in social networks not only affect the order of the platform but also waste platform space.At the same time,it also has a bad impact on users and a bad experience.Therefore,in order to solve the above problems,this paper proposes a spam filtering and recommendation system in social networks.The system proposed in this paper contains two parts.On the one hand,for the spam problem of social networks,because machine learning theory and technology are more mature and have the advantages of high accuracy and low cost,this paper focuses on machine-based garbage.The information filtering technology is compared and analyzed,and the SVM classification algorithm is the best.However,the SVM classification algorithm has long training time and poor adaptability.Therefore,this paper proposes an improved SVM algorithm based on incremental learning.In the incremental learning process of the traditional SVM algorithm,the attenuation factor is introduced to process the SV set.Compared with the traditional SVM incremental learning algorithm,the improved SVM algorithm not only maintains the accuracy but also shortens the learning time.In this paper,the improved SVM algorithm and the traditional SVM algorithm are also verified in the experiment,and the effect in the spam filtering is better.On the other hand,by studying the existing recommendation algorithms and combining the actual needs in social networks,this paper proposes a hybrid recommendation system based on social networks.The system focuses on the user's behavior information in the social network and the meaning of the relationship network to the product recommendation,and at the same time integrates the user-based collaborative filtering algorithm.Firstly,based on the user's scoring matrix,the user-based collaborative filtering algorithm is used to score and predict the products.Then,the social network structure is analyzed,and the random walk algorithm is used to directly focus on the friends and indirectly pay attention to the friends.The score prediction is performed;finally,the weighted manner of the twopredicted scores is combined to obtain the final predicted score.Experiments show that the proposed algorithm can effectively improve the accuracy of the recommendation system.