Research On Personalized Recommendation Of Digital Library Based On Small Data Fusion

With the rapid increase in the number of digital resources,it is more difficult for users to find resources that meet their needs or interests in digital libraries.To deal with this difficult,digital libraries need to provide personalized services and actively recommend content which can meet users' needs.After a lot of research by experts and scholars,the digital library personalized recommendation model has been relatively mature,usually consists of three parts:user information module,user interest module,and recommendation algorithm module.Among them,the user information module is the basis of the personalized recommendation of the digital library,and to a certain extent determines the merits of the recommendation effect.The user information in the existing digital library personalization recommendation research mostly comes from the inside of the digital library management system.The data utilization rate is low,and it is difficult to fully represent user characteristics.With the continuous development of science and technology and electronic devices such as wearable devices and sensors continue to emerge,fine-grained data on individual user behaviors are well documented and collected.These data can more fully and profoundly describe users behavior.This kind of data that records the daily behaviors,hobbies,habits,etc.of users is called "small data",and it is an omni-directional,multi-perspective,fine-grained quantitative data of individual users.Combining small data of users inside and outside digital library systems and electronic devices can describe user characteristics more comprehensively and profoundly.Small data come from a wide variety of sources,with various forms and complex structures.To effectively use the hidden value of small data,it is necessary to organize and integrate it effectively.This article discusses the use of small data fusion methods for digital library personalized recommendation issues,providing a new solution to digital library personalized recommendation research.Data fusion is usually composed of three levels:data layer fusion,feature layer fusion,and decision layer fusion.Based on the actual needs of personalized recommendations for digital libraries,small data fusion can be divided into data layer fusion,feature layer fusion,and knowledge layer fusion.The fusion result of each layer is used as the input of the next layer to continue fusion.In the data layer fusion,the user small data set is used as the input data of the layer,and the small data fusion library is formed through three steps of data mapping,data deduplication,and data update under the guidance of the prior model.The small data fusion library formed after the data layer is merged serves as the input data of the feature layer,and performs attribute analysis,feature extraction,and feature classification to form a small data fusion feature database.After data layer fusion and feature layer fusion,a preliminary characterization of user behavior features was performed.Finally,the small data fusion feature database is input into the knowledge layer fusion,and a small data fusion knowledge base on user needs and preferences is constructed through feature association,feature mining,and situation estimation,so that users are fully and deeply characterized.Through summarizing the general model of personalized recommendation of digital library,this paper discusses the integration of small data fusion method into the process of personalized user recommendation of digital library.After fully analyzing the digital library user types and behavior characteristics,based on the small data fusion knowledge base,the user preferences and user requirements are extracted and the time factors are taken into account to form a user preference view and a user demand view that change over time.The dynamic angle builds the user model.After constructing a complete user model,the overall idea of personalized recommendation is to match the user model with the resources after the organization of the ontologies,that is,to perform personalized recommendation work,and then to feedback the user's evaluation of the recommendation results back to the user model to update the user.model.Research-oriented users include teachers,students,employees,etc.who are engaged in scientific research in colleges and universities.Scientific research personnel's academic level,professional skills,and information literacy are generally higher than those of ordinary users,and their professionalism and timeliness in providing services to digital libraries are also more demanding.This paper takes the research user as an example,introduces the process of personalized recommendation of digital library based on small data fusion from the aspects of research user small data fusion,research user model construction and research user resource recommendation.The research user small data fusion clearly defines the research user's small data source and content,and has a clear understanding of the research user's behavioral trajectory.Based on this,the research user's small data is obtained to provide data foundation for small data fusion.After acquiring user small data,data fusion of user small data is performed and a user model is constructed.In the process of personalized recommendation,the idea of collaborative filtering recommendation algorithm is adopted to perform user-user collaborative filtering and project-item collaborative filtering.Based on the related theories of small data,data fusion and personalized recommendation,this paper proposes a personalized recommendation method for digital library based on small data fusion and takes the research user as an example to introduce how to use this method.This paper provides a new research idea of personalized recommendation of digital library.