| With the advent of the new pattern of the network information, the role of people in the internet gradually changed. On the one hand, as the information browser, people can use more abundant cyber source to meet their own needs. On the other hand, as the information producer, people are accustomed to uploading the little things to the internet, at the same time continue to produce resource with unprecedented speed. The emergence of this mass information makes users feel confused, and very difficult to pick out the interested content, this phenomenon is called information overload. So, to solve the information overload problem has become increasingly urgent.The performance of recommendation system is outstanding in solving the information overload problem, become a hotspot that many scholars pursue. Recommendation system get the original data which can describe users’ interest through users historical behavior log in the server, and then construct the users’ interest model, through the similarity analysis and calculation, a more personalized browsing page can be presented to users, so the users’ satisfaction is improved. Recommendation system is not only a hot academic research algorithm, it is more important has been widely used in the internet as an effective marketing tool. However, recommendation system has exposed some problems in the face of increasingly complex and diverse application scenarios, such as data sparse problem, cold start problem and user drift interest problem etc.In this paper, with movie recommendation and the storage strategy of medical freezer for applications, we propose effective solutions in view of problems existing in the prior art.The main contents are as follows:(1)For the movie recommendation as the application background, we proposed a mixed recommendation algorithm based on users’ interest vector. As everyone knows, Collaborative filtering is the most widely used recommendation technology in the personalized recommendation system and some related fields. However, the traditional collaborative filtering algorithms do not take the change of users’ interests into account, and suffer from the data sparse problem of users’ preference matrix. Therefore the user’s hybrid interest vector is utilized to solve the data sparse problem in this paper, and the corresponding recommendation algorithm is proposed. In the proposed algorithm, the interest vectors of users according to the content of movies and the preference matrix of users are obtained via iteration, the users’ similarity matrix according to the interest vector and users’preference are constructed, and the personalized recommendation is given by using the traditional collaborative filtering methods. In order to track the changes of users’ interests, the time factor is taken into account. The comparison experiments are performed on the Movielens data sets and the results show that the proposed algorithm has better performance in convergence and accuracy.(2)For the medical refrigerator as the application background, we proposed the intelligent access strategy of intelligent medical refrigerator system sample based on users’ behavior. This policy is mainly to solve the following technical problems, ①how to help user more convenient to store and pick sample by using rich sample content information.② how to build effective access strategy based on users’ access behavior. The establishment of users’behavior data is real-time collected by the user’s storage and their extraction, and we establish the correlation matrix between samples through analysis characteristic attribute of the sample and users’ behavior data, finally propose storage location that aim at be stored samples(the problem①), and get the samples storage location distribution relative optimization. At the same time, in the sample extraction stage, we put forward the corresponding recommendations, so as to enhance the users’ experience and improve the efficiency of extraction (the problem ②)。... |
