Based On The Time Features Of The Network Structure Is Recommended

With the vigorous development of social progress, and Internet technology, thenumerous web2.0and web3.0of the website to enter people’s attention, people canfreely publish information on the Internet to purchase goods. In this context, we enteredthe era of information explosion and extreme pursuit of individual, each user in thepursuit of their own independent needs are precisely met the era of huge amounts ofdata. Finding resources and publishingresources have become the most important taskto the most appropriate user. Recommendetion techniques provide a viable way for us tosolve the information overload problem. Recommendation algorithm is relativelydeveloped, to bring the convenience of our lives.Recommendation methods are used with the user’s purchase history information,whose chief representative is collaborative filtering, SVD, and diffuse algorithm. In fact,there are a lot of other information in the purchase information can be used, for example,time, personalized labels. In recent years, a large number of researchers have usedadditional information to improve recommendation effect.In this paper, we add the time factor to diffuse algorithm, to modify the weights ofthe intermediate steps to achieve the effect to improve the accurate rate. At first weintroduce the timing model to analyze the properties of the user and commodity changesover time.Then every step of the algorithm on the basis of the analysis for the massdiffusion time expansion factors, has improved in both accuracy and diversity, andfinally the method is applied to the labeled substance algorithm is extended to threegraph algorithms, and found that it still is effective. In addition, two methods whichchange the structures of graph by adding nodes or edges has improved the accuracy ofthe system in the experiments.