【作者】 周琳；

【导师】 刘东苏；

【作者基本信息】 西安电子科技大学 ， 情报学， 2021， 硕士

【摘要】 近些年,世界的科技水平得到了巨大的提升,移动支付,即时通信,远程控制等新兴技术在为人类活动带去便捷的同时,也提高了人们的生活质量。科技的进步源于知识的不断发展,知识的发展需要学者的不断研究与创新。随着科研学者对推动世界文明进步的重要性日渐突出,对于学者进行相关研究也逐渐增多。学者画像技术作为开放的互联网环境中精准获得学者信息的方式,有着辅助管理者进行管理,协助完成学者多方面信息推荐,促进科学发展的作用,因此,其构建方法也成为各领域的研究热点。在学者画像的构建过程中,除了对学者进行多维度信息的标签展示外,学者的学术影响力大小以及合作关系发现也是研究的热点内容。研究学者的学术影响力,可以对学者的能力做出准确评价,有助于精准定位所需要的学者人群,而发现学者潜在的合作关系,对加深学者之间的交流与知识再创造也有着积极影响。因此本文首先构建多源数据融合的学者数据集,对学者画像进行多维度构建,并在此基础上应用画像中的相关指标数据,对学者的学术影响力以及潜在合作关系的发现做出研究。基于本文对学者画像的主要研究方向,做出的主要工作如下:(1)展现一个学者的特征需要多方面的信息,本文在明确学者画像构建维度的基础上,使用爬虫等技术对百度百科,百度学术等多个网站中的数据进行爬取与搜集,并融合多源异构数据形成学者相关信息的数据集,保证学者信息的多样性。在此基础上,文章针对于本文学者画像的构建维度对学者的标签信息进行发现。使用标签化的形式对学者的相关信息进行完整呈现,并完成对其画像的构建。(2)本文使用画像中的相关信息数据,对学者的学术影响力进行深入研究。不同于以往对学术影响力的研究只针对某一个特定方面的情况,本文将对学术影响力的评价与预测的研究进行结合,提出基于加权模糊逻辑的特定学者群体学术影响力预测的方法,对学者的学术能力进行全面探索。该方法首先融合多指标数据,使用加权模糊逻辑,对学者的学术影响力进行界定,有效的解决了学术影响力评价中研究指标单一与界定标准模糊的问题,接着在发现具有特定学术能力学者的基础上,使用机器学习的方法对学者的学术影响力进行预测,充分发挥机器学习精确性的特点。(3)本文提出基于相似合作能力的潜在合作关系发现的方法,通过合作能力的相似度初步发现与目标学者具有合作可能性的对象,随后引入研究领域相似度与性别同质性指标,完成与目标学者合作可能性的判断,最终通过可能性排名发现潜在合作对象,保证了发现潜在对象的合理性与新颖性,对学者合作关系推荐具有一定的参考价值。更多还原

【Abstract】 In recent years,the level of science and technology in the world has been greatly improved.Emerging technologies such as mobile payment,instant communication and remote control not only bring convenience to human activities,but also improve people’s quality of life.The progress of science and technology comes from the continuous development of knowledge,which requires the continuous research and innovation of scholars.As the importance of scientific research scholars in promoting the progress of world civilization is becoming more and more prominent,the research on scholars is also increasing gradually.As a way to accurately obtain scholars’ information in the open Internet environment,scholar portrait technology has the role of assisting managers in management,assisting scholars in various aspects of information recommendation and promoting scientific development.Therefore,its construction method has become a research hotspot in various fields.In the process of constructing scholar portraits,in addition to the multi-dimensional label display of scholars,the academic influence of scholars and the discovery of cooperative relationships are also the hot topics of research.Studying the academic influence of scholars can make an accurate evaluation of the abilities of scholars,which helps to accurately locate the required group of scholars.And discovering the potential cooperative relationship among scholars also has a positive impact on deepening the communication and knowledge recreation among scholars.Therefore,this paper first builds the scholar data set with multi-source data fusion,constructs the multi-dimensional construction of the scholar portrait,and applies the relevant index data in the portrait on this basis to study the academic influence of scholars and the discovery of potential cooperation relationship.Based on the main research direction of scholars’ portraits in this paper,the main work is as follows:(1)Show the characteristic of a scholar to various information,based on the clear scholar building dimensions,on the basis of the use of reptiles and other technology,Baidu encyclopedia academic data in multiple sites such as Baidu crawl and gathering,and fusion of multi-source heterogeneous data form scholars related information data sets,to ensure the diversity of academic information.On this basis,this paper aims at the construction dimension of scholars’ portraits to discover the label information of scholars.The tagging form is used to present the relevant information of the scholar completely,and the construction of his portrait is completed.(2)This paper uses the relevant information data in the portraits to conduct an in-depth study on the academic influence of scholars.Different from the previous research on academic influence that only focuses on a specific aspect,this paper combines the evaluation and prediction of academic influence,proposes a prediction method of academic influence of a specific group of scholars based on weighted fuzzy logic,and comprehensively explores the academic ability of scholars.This method first fusion more index data,using the weighted fuzzy logic,to define the academic influence of scholars,effectively solve the academic research on the evaluation of the impact indicators of the single and define the standard fuzzy,then on the basis of academic ability scholars found a specific,using machine learning method to predict the academic influence of scholars,Give full play to the accuracy of machine learning.(3)In this paper,based on the ability to work in a similar method,found on the potential partnership through preliminary cooperation ability of similarity found that scholars with possibility of cooperation with the target object,and then introduced similarity and gender research field homogeneity index,complete with the target scholars cooperation possibility judgment,finally through the likelihood ranking find potential partners,It ensures the rationality and novelty of the potential object,and has a certain reference value for the recommendation of cooperation between scholars.更多还原