請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51598
標題: | 利用傳統和非傳統文獻計量指標的組合,預測學術期刊論文影響力 Predicting the Impact of Academic Journal Article through a Combination of Traditional and Non-Traditional Bibliometric Indices |
作者: | Ting Hsiung 熊庭 |
指導教授: | 劉鋼(Kang-Ernest Liu) |
關鍵字: | 替代性計量,關鍵字熱門程度,預測引用,機器學習,隱含狄利克雷分布, Altmetric,Keyword Popularity,Predictive citation,Machine learning,Latent Dirichlet Allocation, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 如何以客觀且效率的方式評斷一篇論文的品質與影響力,一直是學術界廣為討論以及研究的重要議題,透過預測論文的引用次數,我們可以評估論文作者在學術上的未來影響力,幫助研究人員找到具有影響力的文章,並且有效理解研究議題。 隨著網路蓬勃發展,人們開始使用網路媒介交流科學研究,若只使用傳統文獻計量來預測學術影響力,將可能忽略網路媒介所帶來的影響,進而產生偏差。故本研究使用傳統文獻計量指標:引用特徵、期刊特徵、作者特徵、論文特徵四大特徵,以及科學出版物在各種網路平台(新聞、部落格和社交媒體等)上獲得關注的量化指標Altmetrics與期刊文章關鍵字熱門程度等非傳統文獻計量指標來預測期刊論文的未來影響力。旨在建立傳統文獻計量指標與非傳統文獻計量指標來預測高引用期刊文章,最後使用六種機器學習分類模型:K-近鄰演算法、支援向量機、樸素貝氏、隨機森林、神經網路、極限梯度提升,找出最佳分類模型。 本研究實證發現,行銷以及旅遊領域之期刊文章在加入非傳統文獻計量指標後,皆會提升預測性能,且六種分類模型中,xgboost得到最佳的高引用文章的識別能力。基於以上實證結果,能夠幫助研究人更精準地識別出未來具有高影響力的文章。 How to judge the quality and influence of a paper in an objective and efficient way has always been an important topic in academia. We can evaluate the future influence of the paper by predicting citations. It can help researchers find influential articles and understand effectively of the research topics. As the Internet flourishes, people communicate scientific research with internet. If only predict the Impact of Academic Journal Article through traditional bibliometrics indices, it may be biased by ignoring the influence of the Internet.Therefore, this study uses traditional bibliometric Indices: Citation characteristics, Journal characteristics, Author characteristics, and Paper characteristics; non-traditional bibliometric such as Altmetrics that’s the quantitative indicators of scientific publications gain attention on various online platforms (news, blogs, and social media, etc.) and Keyword popularity to predict the influence of journal articles. The main purpose of the thesis is to predict the highly cited journal articles through a Combination of Traditional and Non-Traditional Bibliometric Indices. Using six machine learning classification models: K-nearest neighbor algorithm, support vector machine, naive Bayesian, random forest, neural network, xgboost to find the best classification model. The empirical results show that the addition of non-traditional bibliometric indicators to journal articles in marketing and tourism fields will improve prediction performance. Among the six classification models, xgboost has the best ability to identify highly cited articles. Based on the above empirical results, it can help researchers to more accurately identify high-impact articles. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/51598 |
DOI: | 10.6342/NTU202002730 |
全文授權: | 有償授權 |
顯示於系所單位: | 農業經濟學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-0908202021374700.pdf 目前未授權公開取用 | 5.45 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。