Scholar Impact Prediction Based On Collaboration Patterns

With the continuous development of academic society,the output of academic papers has been increasing in recent years.The academic information has been continuously enriched.Scholarly big data has gradually become a new research field.Scholar impact evaluation is an indispensable part of academic society.In the academic community,we can choose to track the influential scholars to getto know the latest research progress of a given research area;in the college position assessment,we can take advantages of scholar’s academic influence to determine the job evaluation;in the fund application,highly influential scholars can be more likely to receive funding.However,with the advent of the era of large data,academic information overload problem has gradually become a barrier to academic impact evaluation.That is,it is difficult for scholars to evaluate the influence of a scholar.Therefore,how to effectively obtain and predict the influence of a scholar is an urgent problem to be solved.In order to solve the problems and challenges posed by the large number of academic journals,researchers have studied and analyzed various academic relations from different perspectives.In this paper,we mainly study the scholar impact prediction.In other words,given a scholar,we try to predict his/her academic influence in the future.This work will help to understand the scholar’s research ability and impartial evaluation of a scholar’s academic ability,which will shed light on fund distribution,job evaluation and so on to provide practical problems to solve the problem of academic information overloadAt the same time,deep learning has become the research hot topic of machine learning in recent years.The core idea of deep learning is through the analysis and simulation of the human brain’s data processing methods,processing the data by imitating human neural network thinking.In the information age with huge amount of data,the algorithm of deep learning has the potential in prediction systems.At the same time,we propose a method to predict the academic influence based on the collaboration pattern of the scholar,which is totally different with traditional citation-based prediction system.Different from the traditional prediction method,we use the deep learning method for scholarimpact prediction.We introduce the deep learning method in the traditional prediction model.In this paper,we propose a new model called SIP(Scholar Impact Prediction),which includes two stages: the academic collaboration network based collaboration pattern feature extraction,and the impact prediction based on stacked autoencoder.First,collaboration pattern including local and network features are extracted from the scientific collaboration networks as the input features.And then,stacked autoencoder algorithm is employed to predict scholar impact.In this paper,we evaluate the performance of SIP method on the DBLP data set,which shows that our proposed model can generate more accurate results compared with traditional machine learning methods in terms of MAR,RSME,and PCC.It means that deep learning can solve such issues.In addition,we analyze the influence of each input feature on scholar impact prediction,which can help understand the mechanisms of academic impact evaluation.On the other hand,this paper explores the problem of scholar impact prediction in the academic society.In real life,in the online social network people also have the same problem.Because the network properties of different prediction techniques are different,but they are interlinked with each other.Therefore,we have proposed the prediction method,which not only has a certain universality,but also can be applied to other prediction and recommendation systems.For example,it can be used in items recommendation in e-commerce recommendation,online social media,as well as recommended.Therefore,the proposed prediction strategy has certain universality,and it has certain reference significance to other prediction technology.