Research Of Data Recommendation For Earth System Science Data Sharing

Earth system science is a typical data-intensive discipline.It mainly studies the laws and mechanisms in the interaction between the various layers and subsystems of the Earth.Disciplinary related research requires the use of interdisciplinary,cross-regional,and multi-scale integrated data in the field of Earth system science.Data sharing can be used for effective data mining and utilization,and promote academic exchanges and cooperation.With the advent of geosciences big data era and new geographic information era,the development of information acquisition technology has made the subject-related research data show an explosive growth trend.At the same time,the development of Internet technology has enabled more and more Internet users to participate in the Earth system.Scientific data sharing and other geo-information related cyberspace information services.The increase of data volume and the expansion of user groups have caused the excessive amount of data in the Earth System Science sharing process to exceed the user's acceptable level.It is difficult for users to find the information and data they need in the massive data,and there is "data overload".problem.Therefore,in the process of Earth system science data sharing,research on how to accurately and efficiently obtain data and design data recommendation methods is of great significance to both professional researchers and the general public.Designing recommended methods for earth system scientific data requires the following key issues to be solved:First.earth system scientific data has the characteristics of large data volume,heterogeneous structure,and complex information,and needs to be recommended when using traditional recommendation methods for data recommendation.The characteristics of the objects are described in a unified manner,and there are difficulties in describing the scientific data of the earth system in a structurally isomorphic manner.Secondly,the relationship between Earth system scientific data is complex,and the user's needs are diverse.A single recommendation method based on data attributes or user behavior is difficult to meet the user's specific requirements for data,and a recommendation method needs to be designed from multiple perspectives.Finally,there are many users of Earth System Science data sharing platforms,and the amount of data they contain is huge.The vast majority of user feedback information is implicit feedback,and does not include the specific attitude of users to the data.The user-data relationship table established based on these feedback information is included in the table.The characteristics of high-dimensional,sparse,heterogeneous and containing redundant information and noise are presented.It is necessary to study how to use these user feedback information to conduct data recommendation research based on the characteristics of the discipline.Based on the above issues,this paper starts with two aspects of data characteristics and user needs,and conducts research on data recommendation in the process of Earth System Science data sharing.The following major results have been achieved:1.Data-oriented recommendation describes the characteristics of earth system scientific data.Earth system science data are massive,heterogeneous,and complex and are not suitable for characterizing using a single isomorphism.This paper analyzes the characteristics of the Earth system scientific data and the specific needs of users in the research process.Based on the metadata standard,the key information describing the characteristics of the data is extracted from the metadata of the shared data and expressed hierarchically from the data itself.The angle describes the characteristics of the Earth system science data and provides necessary support for the design of content-based recommendation methods.2.Use the user's implicit feedback to cluster the data.Based on the historical behavior data of users,the concept of complex networks is introduced to analyze the characteristics of data sharing networks,and a method of using implicit feedback information to obtain similarities between data is designed.Based on this,the research data is clustered according to research topics.From the user's point of view,describe the features of the Earth system scientific data,and provide the data foundation for the design of recommendations based on research topics.3.Use the characteristics of the data itself and the research topics summarized in the user feedback to jointly recommend.By comparing the similarities and differences between recommendation of scientific data in the earth system and e-commerce domain,based on the feature extraction of data content and clustering of research topics based on user feedback information,a recommendation method based on data features and research topics is designed.Compare the performance of the two methods to achieve a combination of the two methods.Based on the above research results,the data of the Yangtze River Delta Scientific Data Center,a national data platform for the sharing of Earth system science,was selected as the research object,and the recommendation test standards were proposed and data recommendation tests were conducted.The test results prove the feasibility of the personalized recommendation system of the Earth system scientific data,and prove that the personalized recommendation algorithm for scientific data of the Earth system proposed in this paper can effectively recommend the data according to the user's preferences during the scientific data sharing of the earth system.To ameliorate the problem of "data overload" existing in Earth system science data sharing,and provide reference for other related research.