Research On Collaborative Filtering Algorithms Based On User Time And Trust

With the continuous advancement of globalization in the world,the Internet occupies an indispensable position in our daily life.More and more information is filled with all aspects of our lives.Finding useful resources from the vast amount of information and providing convenient information for users are still the unremitting goal of every computer person.In such a trend,the recommendation system will play its due function.It can provide products of interest to the target,and then recommend them.It is more convenient for users to buy,and it also greatly reduces the time problem of searching items and improves the efficiency of searching.The recommendation system not only records the type,time and mode of items purchased by users before,but also analyses and integrates the items purchased by users.It also designs models,algorithms and directional recommendation for users.Therefore,the recommendation algorithm is particularly important,and it is related to whether the user's model is established accurately and whether the recommended products can meet the user's needs.Collaborative filtering algorithm mainly promotes users to purchase similar target products by recommending items that they are interested in but not purchased.Traditional collaborative filtering algorithms have many drawbacks:too many users,high cost of calculating score matrix;being not able reflect users' current interests and hobbies in a timely manner;data sparsity,cold start problem when recommending new projects,being not able reflect users' interest changes,and so on.In order to solve a series of problems in the process of recommendation,three improved filtering algorithms are proposed to improve the accuracy of recommendation.Firstly,a time-varying model is proposed.The time value is added into the algorithm process to effectively distinguish the long-term interest from the short-term interest of users.According to the time of interest change,the model of user interest changing with time is constructed.According to different weights,the model is added to the collaborative filtering algorithm to better monitor the recent interest change of users and recommend commodities for users.The recommendationeffect is more accurate when the time-varying value is added clearly.Secondly,a trust model is proposed.Considering the degree of user trust,the user trust is added to the algorithm to establish a trust model,which can measure the degree of user trust,and then distinguish whether the user's rating data can be trusted,and the proportion of the situation,and predict the item rating value,so as to recommend goods for users.Finally,a hybrid model is proposed,which will be timely.Considering both inter factor and trust factor,building a model can combine the advantages of the two and join into the algorithm to form a hybrid collaborative filtering algorithm,which can accurately recommend to users.Experiments show that the improved algorithm is more accurate.