隨著典型相關分析應用的日益廣泛，線性典型相關分析並不足以描述兩個主題事務間確實的關連及與其個別相關事務的互動關係，為了更適切地解釋資料數據背後所表達的現象，近來不斷地有關於非線性典型相關分析的理論發表，本文利用1992年Sheng所發表的非線性典型相關分析的理論，將之應用在對臺灣地區民國八十二年的前二十五名最有聲望的企業的資料分析上，並比較應用線性典型相關分析與非線性典型相關分析做資料分析的結果。

本論文對線性三對角方程組之解法提出平行演算法於超立方體網路 ( hypercube network), 並且此平行演算法能達到最佳費用 (optimal cost ) O(N). 討論的解法包含 (1)循環消減法 (cyclic reduction method)及 (2)高斯消去法 (Gaussian elimination method), 基於 (1)法之平行演算法當使用處理器個數為 O(N/logN)時, 其執行時間為 O( logN); 基於 (2) 法之平行演算法當使用處理器個數為 O(N/(logN)^2) 時, 其執行時間為 O((logN)^2); 費用 (cost) 等於處理器個數乘以執行 時間.

雲端運算和資料採礦已成為這二十一世紀的重要發展方向，綜觀現今各個生活層面，已漸漸的融合雲端計算的技術，故結合雲端運算已是一種趨勢。簡而談之，雲端運算是一種讓使用者更加地快速、便利又省成本的一種技術。而資料採礦方面，也已從先前的專門挖掘數字型態的資料，到現在多元的挖掘，像是文字、圖像採礦。資料採礦雖然比雲端運算發展的早，但是其功用是可以相輔相成的，有鑑於此，本研究係要發展出一資料採礦分析系統，使得使用者方便又簡易的操作。並針對特定的資料採礦分析方法-關聯規則及集群分析去研究，並利用Apriori 演算法及K-means方法，和Microsoft Excel VBA和R軟體共同結合出此資料採礦系統。

Hamilton (1989)發展出馬可夫轉換模型，假設模型母體參數會隨某一無法觀察得到的狀態變數變動而改變，並用馬可夫鏈的機制來掌控狀態間切換，可適當掌握金融與經濟變數所面臨的結構改變，因此是一個十分重要的財務模型。Schwert (1989)觀察股價波動狀況，發現經濟衰退期的股價波動比經濟擴張期大，因此認為Hamilton (1989)所提出的馬可夫轉換模型亦可應用於股票市場。然而，發現當市場上有重大訊息來臨時，大部分標的資產報酬率會產生跳躍現象，因此汪昱頡 (2008)提出跳躍風險下馬可夫轉換模型，以改善馬可夫模型所無法反映之股價不正常跳躍現象。在探討股價指數報酬率之敘述統計量與動態圖後，本文認為跳躍幅度也會受狀態影響，因此進一步拓展周家伃 (2010)跳躍獨立風險下狀態轉換模型，期望對股市報酬率動態過程提供更佳的分析。實證部分使用1999到2010年的國際股價指數之S&P500、道瓊工業指數與日經225三檔作為研究資料，來說明股價指數具有狀態轉換及跳躍的現象，並利用EM(Expectation Maximization)演算法來估計模型的參數，以SEM(Supplemented Expectation Maximization )演算法估計參數的標準差，且使用概似比(Likelihood ratio)檢定結果顯示跳躍相關風險下狀態轉換模型比跳躍獨立風險下狀態轉換模型更適合描述股價指數報酬率。最後，驗證跳躍相關風險下狀態轉換模型能捕捉其報酬率不對稱、高狹峰與波動聚集之特性。 / Hamilton (1989) proposed Markov switching models to suppose the model parameters change with unobserved state variables which control the switch between states by Markov chain. It can be appropriate to grasp the financial and economic variables which facing structural changes, so it’s a very important financial model. Schwert (1989) observed stock prices, and discovered that the volatilities of recession are higher than the volatilities of expansion. Hence, Schwert (1989) suggested to apply the Markov switching models to stock market. However, most of underlying asset return have jump phenomenon when abnormal events occur to financial market. Wong (2008) proposed Markov switching models with jump risks to improve Markov switching models which can not capture the jump risk of asset price. According to stock index return’s descriptive statistics and dynamic graph, we argue that states will impact jump sizes. In this paper, we extend the regime-switching model with independent jump risks (Chou, 2010) to provide better analysis for the dynamic of return. This paper use stock indices of the study period from 1999 to 2010 to estimate the parameters of the model and variance of parameter estimators by Expectation-Maximization (EM) algorithm and SEM(Supplemented Expectation Maximization ) , respectively. And use the likelihood ratio statistics to test which model is appropriate.Finally, the empirical results show that regime-switching model with jump sizes dependency risk can capture leptokurtic feature of the asset return distribution and volatility clustering phenomenon.

給定兩個條件機率分配，判斷他們是否相容？是否有唯一的聯合機率分配？以及相容時，如何找出所有可能的聯合機率分配？是研究相容性相當重要的課題。本文針對有限離散型二維條件機率分配，以Arnold and Press(1989) 最先提出的比值矩陣法，及由Song , Li, Chen, Jiang, and Kuo (2010) 所提出的檢驗法為架構，提出新演算法且利用此演算法來設計程式，使程式能判斷兩條件機率分配是否相容，以及相容後可求出對應的所有聯合機率分配。本文亦依據新演算法並應用MATLAB軟體設計程式，讓使用者可以很快地對上述三個問題得到答案。 / When two conditional distributions are given, the following three important questions are likely to be raised. Are they compatible? Is the corresponding joint distribution unique if they are compatible? How do you find all the corresponding joint distributions if they are compatible? In this thesis, basing on ratio matrix method given first by Arnold and Press (1989), and on the method for checking compatibility existence, for checking uniqueness, and for finding all possible joint distributions provided by Song, Li, Chen, Jiang, and Kuo (2010), we provide a new algorithm to answer these questions. Using this new algorithm, we also provide a MATLAB computer program so that any user could get the answer quickly for the above three questions.

因為疾病空間模式通常會與環境中的危險因子有很強烈的關聯性，因此流行病學家與社會大眾都對疾病的空間模式感到興趣。舉例來說，空間群聚就是一項非常受到重視的疾病空間模式，在眾多的空間群聚檢定方法種，Kulldorff和 Nagarwalla在1995年提出的空間掃描統計量是相當受到廣泛應用的方法，雖然這個統計方法可以檢定初空間資料的異質性，但是卻沒有辦法區隔這些異質性是來自於整體空間資料的相關性或是局部的空間群聚。在本篇論文中，我們將分別提出計次型的統計方法與貝氏統計方法兩種類型的空間群聚檢定方法來處理這樣的問題，其中計次型的統計方法為一兩階段的統計方法，首先採用EM演算法來估計空間自相關，並根據估計的結果與掃描窗格在偵測空間群聚；另一方面，貝氏方法則考慮加入群聚的中心位置及半徑作為事前的機率分布，進而透過MCMC的方法來計算出後驗分布的結果。除此之外，北卡羅來納的嬰兒猝死症和台灣老年人口癌症死亡資料將被用來示範與評價不同群聚檢定方法的差異與效果。

由於資訊科技的進步與網路的普及，企業得以收集與儲存大量的資料。使用資訊工具來協助資料處理、資訊擷取、以及產生知識已然變成企業的重要課題之一，所以如何良好運用資料探勘工具成為使用者關注的焦點。由於並非每一個使用者對於資料探勘的原理都有充分的了解，所以如何從探勘工具提供的功能中選用最佳的解決方案並不容易。如果對於探勘結果不滿意而需要調整軟體邏輯，與IT人員的協商溝通卻又曠日費時。 為了解決這個問題，本研究提出一個演算法選擇方法，藉由分析商業應用的內容，來自動對應到特定的資料探勘方法與演算法，讓選擇演算法的過程更為快速、更系統化，提升利用資料探勘工具解決商業問題的效率。 / Due to the information technology improvement and the growth of internet, companies are able to collect and to store huge amount of data. Using data mining technology to aid the data processing, information retrieval and knowledge generation process has become one of the critical missions to enterprise, so how to use data mining tools properly is users’ concern. Since not every user completely understand the theory of data mining, choosing the best solution from the functions which data mining tools provides is not easy. If user is not satisfied with the outcome of mining, communication with IT employees to adjust the software costs lots of time. To solve this problem, a selection model of data mining algorithms is proposed. By analyzing the content of business application, user’s requirement will map to certain data mining category and algorithm. This method makes algorithm selection faster and reasonable to improve the efficiency of applying data mining tools to solve business problems.

研究指出，在不同人類個體的DNA序列中，只有0.1%的基因組排列是相異的，而其餘的序列則是相同的。這些相異的基因組排列則被稱為單一核苷酸(SNP)。Affymetrix公司發展出一種DNA晶片技術稱之為Affymetrix GeneChip Mapping 100K SNP set，此晶片可用來決定單一核苷酸資料的基因類型(genotype)。Affymetrix公司採用預設「動態模型演算法」(DM)來決定基因型態。本論文的研究目的是探討與示範對於DM方法中預設的S值的四種修正方式。而這四種修正的方法分別是： (1) Standardized L value，(2) Median-polished L value，(3) Median-center L value，和(4) Median-standardized L value。為了比較S值與四種改進方法，本研究藉由SNP的模擬資料來進行比較。資料的模擬是基於利用改寫過的階層式之Bolstad模型（2004），而模擬模型的參數估計是利用華人細胞株及基因資料庫中95位台灣人的100K SNP資料。根據AA模型與AB模型模擬資料的基因型態正確率，Standardized L value是最好的判斷基因型態之方法。在另一方面，因為DM方法並不是設計來決定Null模型的基因型態，因此對於Null模型模擬資料的基因型態判斷會有問題。關於Null模型的基因型態判斷，本論文提供了一些簡短的討論與建議。然而，依然需要進一步的研究探討。 / It is known there is only 0.1% in the DNA sequences that is different among human beings, and the rest of them are the same. These differences in DNA sequences are defined as SNPs (Single Nucleotide Polymorphism). The Affymetrix, Inc. had developed a DNA chip technology called Affymetrix GeneChip Mapping 100K SNP set for SNP data used to determine the genotype call. The default algorithm applied by Affymetrix, Inc. to decide genotype calls is the Dynamic Model-based (DM) algorithm. This study aimed to investigate and demonstrate four different ways to modify the basic component used in DM algorithm, namely, the S value. These four modified methods include: (1) Standardized L value, (2) Median-polished L value, (3) Median-centered L value, and (4) Median-standardized L value. In order to compare the S value with the four modified L values, a simulation study was conducted. A hierarchical version of Bolstad’s model (2004) was adopted to simulate the SNP Data. The parameters for the simulation model were estimated based on 95 Taiwanese 100K SNPs data from Taiwan Han Chinese Cell and Genome bank. The Standardized L value was proven to be the best method based on the accuracy of the genotype calls determined according to the simulated data of AA model and AB model. On the other hand, the genotype call for simulated data under Null model is problematic since the DM approach is not designed to determine the Null model. We have given some brief discussion and remarks of the genotype call for Null model. However, further research is still needed. /

本研究旨在探討基於基因演算法(genetic algorithm)在同時考量先備知識水平及學習角色異質互補，以及社會互動關係同質因素下，發展之最佳化問題導向網路合作學習分組策略，是否有助於提升問題導向網路合作學習之學習成效、互動關係、團體效能與團體凝聚力。 本研究採用準實驗研究法，以新北市某國小六年級三個不同班級合計83名學生為研究對象，並將三個班級學生隨機分派為採用基因演算法最佳化分組策略的實驗組，以及分別採用隨機分組及學生自行選擇分組策略的控制組一與控制組二，三組學習者皆在問題導向學習系統（Problem-based learning system,簡稱PBL）上進行不同分組策略之問題導向網路合作學習活動。藉由學習成效與團體效能與團體凝聚力評量，以及分析三組學習者在問題導向學習系統上的學習歷程與互動關係，最後再輔以半結構式訪談，以驗證三種不同分組策略在學習成效、互動關係、團體效能與團體凝聚力上的差異。 結果顯示本研究所提出之最佳化問題導向網路合作學習分組策略具有提升學習成效之效益；本研究所提出之最佳化分組策略對於促進問題導向網路合作學習之同儕互動具有正面效益；採用不同問題導向網路合作學習分組策略組別學習者在團體效能與團體凝聚力上具有顯著差異；採用最佳化分組策略組別學習者在問題導向網路合作學習的滿意度接近同意的水準。 / This study aims to explore whether the optimized group formation scheme based on genetic algorithm helps students enhance learning performance, interaction, collective efficacy, and group cohesion in collaborative problem-based learning environment. Factors associated with heterogeneous complementation of students’ prior knowledge levels and learning roles and the homogeneity of social interaction relationship were simultaneously considered in the genetic algorithm-based optimized group formation scheme. In this paper, a quasi-experimental research method is employed to assess the effects of three different group formation schemes on the learning performance, interaction, collective efficacy, and group cohesion in collaborative problem-based learning environment. Eighty-three students in three different sixth-grade classes in an elementary school in New Taipei City were invited to participant in the experiment and were randomly divided into three groups: the experimental group, which adopts genetic algorithm-based optimized group formation scheme, and two control groups, one is randomly grouped; while the other allows students group themselves. Learners in these three groups all use collaborative problem-based learning system (CPBL) to perform collaborative problem-based learning activities. Learning performance, interaction, group efficacy and group cohesion evaluation are applied to analyze the learning process and interaction among learners in these three groups. Finally, a semi-structured interview is supplemented to validate the variation of these three different group formation schemes in learning performance, interaction, group efficacy and cohesion. The result showed that the genetic algorithm-based optimized group formation scheme helps students promote learning performance and provides positive effects on peer interaction in collaborative problem-based learning. Three group learners adopting different group formation schemes for collaborative problem-based learning show significant difference on group effectiveness and group cohesion. The satisfaction of learners adopting genetic algorithm-based optimized group formation scheme for collaborative problem-based learning reached a nearly agreed standard.

轉職對於職涯發展來說，是非常重要的人生課題；而求職者目前在面臨轉職問題時，大多時候顯得手足無措，只能詢問親友的經驗或者憑著直覺找自己有興趣的工作；整個求職的過程就像是拿人生當賭注，運氣不好時即可能賠上美好的未來。 本篇研究使用國內某知名人力銀行的求職者資料，採用資料科學的方式，利用大量求職者的實際轉職資料來做資料分析與探勘，分析轉職高峰期、工作轉換頻率、跨職類轉職、跨產業轉職及轉職與景氣的關係，並使用J48、Naïve Bayesian Classifier、Logistic Regression、Random Forest、AdaBoost和Support Vector Machines這6種分類方法來預測轉職行為。 為了方便呈現實驗結果，本研究使用Google App Engine建立了一個轉職分析查詢系統，透過分析結果可以了解台灣各產業與各職類的轉職趨勢，而轉職預測功能也可以提供給求職者與人資人員做為參考。 / Career transition is important for employees. However, most of job seekers are helpless in decision of career transition. They can only make the decision based on the experience from their friends and family members, or by intuition. The decision of job seeking is like a gamble that may lose a better future when they faced with bad luck. This research tried to analyse and discover the behaviours of job transition from the job seeking data based on the data science approach. The job seeker’s data used in the study was obtained from the well-known job bank’s database. We analyse the behaviours of the job transition, including the peak months of transition, transition frequency, cross-job and cross-industry career transition. Moreover, we investigate the methods to predict the behavior of job transfer. Six kinds of classification algorithms were used to predict the behavior of career transfer, including the J48, Naïve Bayesian Classifier, Logistic Regression, Random Forest, AdaBoost and SVM. We develop the web-based Career Transition Analysis System to provide users the capability for behaviour analysis and prediction of career transition based on Google App Engine. The findings in this study are helpful for industry trends and career transition forecasts for job seeker and human resource staffs.