The Matrix Factorization Recommendation Algorithm Combined By Time Factor And Trust Mechanism

With the rapid development of the Internet and the arrival of the information age,the amount of data is increasing explosively,and the recommendation system has become a more and more important field.Because the recommendation system is based on the user’s historical behavior and the historical characteristics of the project,the user’s preferences and the popularity of the project will change over time,so the time factor will have an impact on the user’s rating behavior,thus affecting the accuracy of the recommendation result;in addition,in the analysis of the user,not all participating users can provide reliable and effective recommendation for the target users,so not all scoring records are valuable for the final recommendation results,only trusted users can generate trust recommendation.Therefore,this study will focus on the above two aspects.First of all,in order to make the final recommendation result reliable,and analyze the influence of the time factor on the user’s trust degree,this study proposes a combination of two trust mechanisms and time factors.Due to the user’s trust level will continue to decline over time,so this study takes time as attenuation factor into the trust computing model.Trust mechanism has two uses in recommendation system: trust filtering and trust weight.This paper will combine these two purposes,make the role of the trust mechanism in the recommendation system can be fully utilized.Secondly,in order to fully analyze the changes of user preferences and project popularity with time,improve the accuracy of recommendation system,this paper proposes the method of integrating the time factor into the traditional matrix decomposition technique,user bias Bu,item deviation Bi and user characteristic vector Pu are respectively as a function of time t.By adjusting the range of time t,the influence of time factor on the accuracy of recommendation is observed.Finally,this paper will respectively test the trust mechanism incorporating time factor and matrix decomposition method on the Netflix data set.By comparing the experimental results with the traditional recommendation algorithm,we analyze the influence of time and trust mechanism on the recommendation accuracy.