Research And Application Of Information Directional Recommendation Technology For Instrument Sharing Platform

At present,with the continuous expansion of the scale of scientific research facilities and instruments,the scope of coverage continues to increase,and the comprehensive benefits have been rapidly improved,but there have also been cases of low utilization rates such as idle instruments and waste.In order to improve the utilization efficiency of instrument resources,the government proposes to build a unified network management platform to realize the organic connection of scientific research facilities and information sharing.In response to the call,the instrument sharing platform was quickly built in many colleges and universities,especially the scientific research facility modules such as appointment,management,supervision and evaluation have been put into use,but the information sharing module has appeared problems of low click rate and little interest.The recommended news information does not meet the user's needs,which has seriously affected the user's experience.In view of the above situation,this paper improves the traditional news recommendation algorithm based on the laboratory project,and designs a crawler and recommendation system,which is applied to the instrument sharing platform.The specific contents include:(1)Algorithm selection: This paper analyzes the implementation principles of several mainstream news recommendation algorithms,studies the applicable scenarios of each algorithm,and finally decides to integrate the trust model and clustering algorithm into the user-based collaborative filtering algorithm,and makes improvements on this basis.(2)The improvement of trust algorithm: There are two shortcomings in the previous trust algorithm.On the one hand,the influence of malicious users is not considered,so the user's trust model is added on this basis,so that the trust value of ordinary users remains unchanged,while the trust value of malicious users is reduced;On the other hand,the traditional trust model does not consider the user's own attribute characteristics,so this paper plans to integrate the user's own characteristics into the trust algorithm,including age,gender,identity,profession,etc.,which can not only make the trust model more comprehensive,but also avoid the problem of cold startup to some extent.In addition,the indirect trust model is added on the basis of the direct trust model,so that there is a trust relationship between the two users who are not directly evaluated.(3)Improvedcommon similarity algorithm: Hot news often affects the calculation of similarity.To solve this problem,the weight of hot news in the formula was adjusted to reduce the influence of hot news.Finally,a new method to measure the relationship between users was obtained by combining the similarity algorithm with the trust model.(4)Use clustering algorithm on user set to achieve good classification effect: The clustering algorithm can make the users of the same cluster more similar,while the user gap between different types becomes larger.Therefore,the use of clustering for collaborative filtering can more efficiently find the users' nearest neighbors,and thus recommend news more in line with users' interests.(5)The crawler module and the recommendation system module are designed on the instrument sharing platform.After testing,good results have been achieved.