Social Network Mining for Learning Research Community Contribution Patterns

Abstract

Accumulation of data from the various sectors has been increasing dramatically over the newlineyears, particularly in the academic and research areas. More research for supporting the research newlinecommunity itself in terms of scanning the domain, measuring the literature relevance, estimating newlinethe prevailing demands and ensuring useful and daunting efforts should be considered. Hence for newlineany research scholar this sort of support would be an added advantage for mechanizing this newlineprocess for significant works connected to current research. To address this problem by applying newlinethe computing procedures available areas like data mining (underlying dataset being relational newlineunnormalized data) and Text Mining (underlying dataset being ordinary text files). As well the newlineSocial Network Analysis (Society meant for authors or research contributors and their prioritized newlinecontributions). The primary dataset used for this context happens to be the collection of instances newlinewith attributes pertaining to collaborations in cross domain literature obtained from various newlineconferences and publications, which is familiar and indexed in ArnetMiner (AMiner) for fifteen newlineyears from 1990 to 2005. newlineThe main part of the research work highlights on the items enumerated as follows. newlineClassification of Research collaborations of an article based on distribution of reference articles newlinefrom academic social network dataset for enhanced accuracy, Attribute Selection (Information newlineGain Ratio) for less time complexity and better accuracy, Meta Classification (Ensemble) for newlineefficient mining of academic research community data, Classification research articles from newlinehighly interconnected domains (with difficulty of associating with single broad domain) with newlineTopic Modeling and Supervised Learning , Identifying the leading research contribution in the newlineclusters of research social networks, namely the top twenty nodes are identified with newlinecombination various weighted schemes. Finally the performance comparisons are made with newlineearlier schemes and established the advantages of the proposed scheme. newlineKey Terms: Data Mining, Academic Social Network, Information Gain Ratio, Meta, newlineTopic Modeling, Individual Frequency, Weighted Frequency. newline newline newline