資訊發達的時代，資訊取得的方式與管道比起以前更方便而多元，但龐大資料量同時也造成了我們往往很難找到真正需要資料的問題，也因此資料的排名(ranking)問題就變得十分重要。本研究目的在於運用排名學習找出良好的排名，利用人對於某特定議題所給予的排名順序找出排名規則，並應用於資料探勘上，讓電腦可自動對資料做評分，產生正確的排序，將有助於資料的搜尋。 　　本研究分為兩部分，第一部份為排名演算法的設計，我們改良現有的排名方法(RankBoost)，設計出另一個新的演算法(RealRankBoost)，並且用LETOR benchmark實測，作為與其他方法的比較和效果提升的證明；第二部份為非對稱加權概念的提出，我們考量排名位置所造成的資料被檢視機率不同，而給予不同的權重，使排名結果能更貼近人類的角度。 / With the innovation in computer technology, we have easier ways to access information. But the huge amount of data also makes it hard for us to find what we really want. This is why ranking is important to us. The central issues of many applications are ranking, such as document retrieval, expert finding, and anti spam. The objective of this thesis is to discover a good ranking function according to specific ranking order of the human perceptions. We employ the learning-to-rank approach to automatically score and generate ranking order that helps data searching. This thesis is divided into two parts. Firstly, we design a new learning-to-rank algorithm named RealRankBoost based on an existing method (RankBoost). We investigate the efficacy of the proposed method by performing comparative analysis using the LETOR benchmark. Secondly, we propose to assign asymmetric weightings for ranking in the sense that incorrect placement of top-ranked items should yield higher penalty. Incorporation of the asymmetric weighting technique will further make our system to mimic human ranking strategy.

本研究旨在建構台灣地區大學排名指標，並藉以評估大學辦學品質。研究方法部分，先以文獻分析歸納出大學排名指標之九大構面與六十八項指標，再以專家問卷以及模糊德菲術問卷進行調查。模糊德菲術調查樣本為二十位高等教育學者與行政首長，本研究透過三角模糊數整合專家對指標重要性之看法並篩選指標項目，最後以歸一化之方式求得各構面以及各項指標權重，完成台灣大學排名指標體系。根據研究之結果與分析，歸納主要結論如下： 一、本研究建構之台灣地區大學排名指標，含九大構面共29項指標。 指標九大構面依權重高低依序為： 教師素質(12.7%)、學校課程與教學 品質(12.5%)、研究表現(11.7%)、大學聲望(11.6%)、學生素質(11.5%)、 學生與畢業校友表現(11.5%)、學校資源(10.0%)、國際化(9.7%)、校園弱勢關懷(8.8%)。 二、教師素質構面共包含三項指標：具博士學位之專任教師比例(4.4%)、專 任教師中教授所佔比例(4.2%)、專任教師比率(4.1%) 三、學校課程與教學品質構面共包含兩項指標：師生比(6.5%)、大學生對大學課程的評價(6.0%) 四、研究表現構面共包含八項指標：全體教師平均獲得研究獎助數(1.5%)、曾獲國家層級學術獎項之教師比率(1.5%)、具全國性專業學會院士成員身分之教師比例(1.5%)、全體教師在Nature、Science刊物，SCI、SSCI、TSSCI、EI以及A&HCI收錄期刊之論文發表平均數(1.4%)、全體教師在Nature、Science刊物，SCI、SSCI、TSSCI、EI以及A&HCI收錄期刊之論文平均被引用次數(1.5%)、全體教師刊載於國內有外審制度期刊與研討會之論文平均數(1.4%)、全體教師發表於國際研討會之論文平均數(1.5%)、全體教師教師專書出版之平均數(1.4%) 五、大學聲望構面共包含三項指標：國內學術同儕聲望調查(4.0%)、雇主對畢業生之滿意度評價(3.8%)、畢業生對母校評價(3.9%) 六、學生素質構面共包含兩項指標：新生甄選入學接受率(4.9%)、以考試分發入學新生之學科測驗平均成績(6.6%) 七、學生與畢業校友表現構面共包含三項指標：五年內學生贏得全國性學術獎項數(3.7%)、該年度畢業生就業(畢業六個月內覓得全職工作)及繼續唸研究所的比例(4.1%)、學以致用率(3.7%) 八、學校資源構面共包含兩項指標：每生之學校年度經費總額平均(4.8%)、每生平均年度學校圖書設備經費(5.2%) 九、國際化構面共包含三項指標：以華文以外領域為主修之國際學生比率(3.2%)、國際教師比率 (3.0%)、全校國際合作計畫件數(3.5%) 十、校園弱勢關懷構面共包含三項指標：招收弱勢學生(2.8%)、大學生平均在校工讀時數(2.7%)、學校年度經費作為清寒學生補助之比例(3.3%) 最後，本研究依研究結果分別提出以下建議： 一、對高等教育主管機關之建議 二、對進行、發布大學排名者之建議 三、對排名資料使用者之建議 四、對未來研究之建議 / The purpose of this study is to construct the Taiwan university ranking indicators which aim to evaluate the education quality of universities. As for research methods, by means of literature review, 68 indicators within 9 main dimensions had been organized as a raw model of Taiwan university ranking indicators based on which the Fuzzy Delphi questionnaire was developed and the survey was conducted with the sample size of 20 higher education experts. Symmetric triangular fuzzy number then was used to analyze experts’ opinion on the importance of each indicator and to help indicator selection. At last stage, normalization of fuzzy number’s total score determined the weight of each dimensions and indicators; accordingly, the Taiwan university ranking indicator system was constructed. The main conclusions are as follows： 1.The Taiwan university ranking indicator system consists with 9 dimensions and 29 indicators in total. The 9 dimensions are: faculty quality(12.7%), curriculum and teaching(12.5%), research(11.7%),reputation(11.6%), student selectivity(11.5%), performance of students and graduates (11.5%), financial resources(10.0%), internationalization(9.7%), inclusiveness(8.8%). 2.The dimension of faculty quality consists with： percent of full-time faculty with top terminal degree(4.4%), percent of full-time faculty as professor(4.2%), percent of full-time faculty(4.1%) 3.The dimension of curriculum and teaching consists with：staff:student ratio (6.5%), student evaluation of course(6.0%) 4.The dimension of research consists with：research grants per academic staff member(1.5%), percent of academic staff member with National Faculty Awards(1.5%), percent of academic staff member with Academy membership (1.5%), publications on Nature, Science, SCI, SSCI, TSSCI, EI and A&HCI per academic staff member (1.4%), citations per article on Nature, Science, SCI, SSCI, TSSCI, EI and A&HCI (1.5%), articles in peer-reviewed journals per academic staff member (1.4%), articles in international conferences per academic staff member (1.5%), publications of book per academic staff member(1.4%) 5.The dimension of student selectivity consists with：Acceptance Rate(4.9%), Entry score(6.6%) 6.The dimension of reputation consists with：peer assessment(4.0%), employer assessment(3.8%), graduate assessment(3.9%) 7.The dimension of performance of students and graduates consists with：the success of the student body at winning national academic awards within 5 years(3.7%), graduate employment(4.1%), correspondent (3.7%) 8.The dimension of financial resources consists with：revenue per student(4.8%), library spent per student 9.The dimension of internationalization consists with：percent of international students (excludes those who major in Chinese) (3.2%), percent of international academic staff member (3.0%), international cooperation projects(3.5%) 10.The dimension of internationalization consists with：attract students from underrepresented groups(2.8%), working hours at school per student (2.7%), expense as subvene for the poor students(3.3%) According to the conclusions, some suggestions had been proposed： 1.suggestions for higher education administrators 2.suggestions for those who are going to conduct university rankings 3.suggestions for university ranking information users 4.suggestions for further study.

霍奇排名是在近幾年才運用在排名的一種方法。在大多數現在的資料庫 中,資料庫很龐大,有些甚至會需要網路連結,而且很多會有資料不完整或 是資料不平衡的狀況。我們選擇用霍奇排名這種排名方法來處理可能會遇到 的這些困擾。 這篇論文主要目的是想用運用基本的線性代數來研究霍奇排名和推導組合霍奇理論。 / HodgeRank is a method of ranking that is new in recent years. In most of modern datasets, the amount of data is very large, some also need the internet connection, and plenty of them have the feature that incomplete or imbalanced. We use the method of HodgeRank to deal with these difficulties. This thesis is primary using elementary linear algebra to survey HodgeRank and deduce the combinatorial Hodge Theorem.

潛水是網路社群中的普遍行為，並且潛水者常為網路社群中的多數，通常潛水者從社群中獲取得多，但卻貢獻得少，雖然對於整體社群無害，但對於網路社群的貢獻卻相當有限，無助於整體社群的發展與成長。因此，如何激發潛水者更積極參與互動討論，樂於貢獻一己之力，對於網路社群的發展甚為關鍵。特別是在數位學習環境中，更應該積極發展有效激發潛水者策略，以促進潛水者更積極參與社群互動討論的意願，提昇整體社群合作學習動力。而透過讓潛水者感受到自己參與社群互動的重要，提昇潛水者的社會知覺，是否有助於激發潛水者表現出更積極的互動行為，值得進行深入的探討。 因此，本研究基於提昇潛水者的社會知覺，於問題導向學習環境中發展「社會互動排名」與「學習夥伴推薦」激勵機制，以探究其對於激發潛水者在社群互動之「討論區與訊息區的文章張貼篇數及內容層次」、「四階段問題導向學習閱讀心得寫作成效」，以及學習社群中的「網路密度」、「網路直徑」、「中心度」。除此之外，也探究「外向－內向」、「人際和諧－人際問題」、「信任感－迫害感」等基本人格特質，是否與潛水者被激發與否的成效有關，進而歸納激發潛水者的具體有效策略。 研究結果顯示，具「社會互動排名」與「學習夥伴推薦」激勵機制之問題導向學習平台，對於提昇社群討論互動以及學習成效具有正向顯著效益；激勵機制確實能有效激發潛水者，降低潛水情形，並且實施激勵機制對於凝聚整體學習社群網絡亦具有正向的效用。 / Lurking is a common behavior in the network community, and lurkers often take the majority in the network community. They often get more from the community, but give less to it. To the whole community, although it doesn’t do any harm, the contribution they make to the network community is so limited, which can’t help the development and growth of the entire community. Therefore, it is quite crucial for the development of the network community about how to motivate the lurkers to participate in the interactive discussion and contribute to the community actively. Especially in the digital learning environment, it should actively develop the strategies to motivate the lurkers effectively, so as to promote the willingness of the lurkers to participate in the interactive discussion of the community more actively. In this way, it can improve the driving force of the cooperative learning in the community. It deserves deep exploration about whether it can help to motivate the lurkers to present more active behaviors in the interaction by making them feel important to participate in the community interaction and improving their social awareness. Therefore, based on the purpose of improving the social awareness of the lurkers, this study develops the motivation mechanism of “social interaction ranking” and ” learning partner recommendation” in the learning-oriented environment to explore the effects of motivating the lurkers in the community interaction, such as “the number and content levels of the articles posted in the forum and bulletin board”, “writing effects of the four-stage problem-based learning and reading”, as well as the “network density”, “network diameter” and “concentration” in the learning community. Besides, it also discusses whether the basic personality is correlated to the effect of motivating the lurkers, including “introversion-extraversion”, “interpersonal problems- interpersonal harmony”, “sense of persecution- sense of trust”, so as to further summarize the concrete and effective strategies of motivating the lurkers. The study results show the problem-based learning platform with the motivation mechanism of “social interaction ranking” and ” learning partner recommendation” show positive and significant benefits to improve the social discussion interaction and learning effect. Moreover, the motivation mechanism system is proven to motivate the lurkers and reduce the lurking situation effectively, and the practice of the motivation mechanism system has positive effect on the cohesion of the whole learning community.

大型股過去一直被認為平均報酬率低於小型股，但如果從個股來看，不少大型股的績效並不會比指數差。考慮到一般非專業投資人在投資股票時，選擇大型股還是比小型股容易，本論文試圖建構一套在實務上較可行的大型個股選股策略—選擇市值第二大的股票，並定期調整個股。我們以美股標準普爾500指數中前兩大市值的股票，分為兩種投資組合做比較，結果發現，市值最大的股票不容易創造超額報酬，市值第二大的股票，反而締造極佳的超額報酬，此現象在過去3年、5年、10年，尤其較過去20年更為明顯。原因在於市值排名第二的股票，多半屬於排名仍在持續上升的成長股，這些個股基本面尚未到達頂點，故股價還會反應一段時間的基本面利多，採取類似動能策略(Momentum Strategy)的方法，報酬率容易超越指數；市值最大者則因為基本面普遍伴隨市值排名已經到頂，加上投資人對於排名第一的股票，多半易產生定錨效應(Anchoring Effect)，即認為股價可能已經反應其該有的價值，較難創造超額報酬，傾向賣出。故同樣投資大型股，選擇市值第二名的股票會優於第一名。 / According to The Size Effect Theory, small cap securities generally generate greater returns than those of large cap companies. However, this trend has involved into the difficulties of stock picking due to the large number of small caps. In this paper I propose a strategy against the size effect theory, “Second is Better”, to pick the second largest market value security as the single stock investment. I examine the performances of the No.1 and the No.2 largest market cap stocks in the S&P500 and apply a 6-month rebalance to construct two different portfolios, which is similar to the concept of Momentum Strategy that buy the past winners and sell the past losers. I find the No.2 stock outperforms than No.1 stock and generate amazing excess returns in the near mid-to-long-term periods. Because No.1 stocks are more likely to experience Momentum Crash than No.2 stocks due to investor’s anchoring bias as they believe the No.1 stock might have been peaked. No.2 stocks are usually in the growing stages that many investors believe the 2nd largest caps still yet to peak during market value expansion.