Research On Personalized Recommendation Algorithm Based On Trust Network

With the rapid development of Internet application system,Web services have become the most important computing resources and software assets in the Internet.Aiming at the problem of trusted service recommendation and personalized selection in distributed environment,achieving the service was discovered on demand.It has become the key problem and the important link of service-oriented computing.At present,the most common solution is information retrieval system and personalized recommendation system.The information retrieval system to obtain the feedback information by keyword tend to be consistent.Due to all kinds of information and the ways of communication is diversified,internet users are also personalized in their use of information,so the user's complex requirements for information usually cannot be satisfied.In addition,the information retrieval system includes a large amount of information that Internet users do not need,the efficiency of data acquisition is low.Personalized recommendation system can be based on the interests of Internet users to recommend the Internet information to users,and then make the Internet users clear their own preferences.Users not only enjoy the personalized recommendation services provided by the system,but also generate dependency on personalized systems.This thesis introduces the main idea of traditional recommendation algorithm.The concept of trust network is introduced into the recommendation algorithm,based on the user's trust data and the historical score data,we can get the credible set of recommendation user.By using the weighting model of user's characteristics,we can meliorate the unilateralism such as only use of similarity recommendation model.The slope one algorithm is analyzed deeply,and the improved slope one algorithm is used to get the final score.The algorithm proposed in this thesis is more efficient,and has been improved in terms of MAE,accuracy and recall rate.The specific studies are as follows:(1)A weighted slope one recommendation algorithm based on trust network is proposed.First of all,the trust relationship is calculated to obtain the trust space.Then we applied trust recommendation algorithm to obtain the trusted set of recommendation information.Finally,the score of the trust network and the score of the weighted slope one algorithm are combined together to obtain the final score,using the Top-N method to complete the recommendation.Through the simulation experiment,the MAE,accuracy and recall rate of the proposed algorithm are enhanced by comparing the MAE,accuracy and recall rate between the slope one recommendation algorithm based on trust network and the classic recommendationalgorithm.The recommended effect has been improved significantly.(2)A hybrid recommendation algorithm based on user attributes clustering is proposed.In the aspect of individual choice,the problem that satisfies the user's interest and behavioral requirements is modeled as a multiple attribute decision making problem.Aiming at the limitation of the one-sided use of similarity recommendation model,this thesis proposes a weighted model based on user attributes.In a more practical way to overcome the one-sided use of similarity recommendation model.To provide an important basis for the selection of high-quality goods.To ensure the effective reuse of the recommended services.The experimental results show that the improved algorithm is superior to the traditional collaborative filtering algorithm in terms of MAE,accuracy and recall rate.The execution efficiency is higher.Algorithm application is of great significance to improve the efficiency of software and business platform recommendation.