研究顯示，家族企業之數量與經濟影響力在全球企業環境中迅速成長並占有一席之地，成為具競爭力的存在，本研究以我國2001至2008年的上市（櫃）公司為樣本，探討財務分析師針對家族與非家族企業在預測行為上之差異，本研究之迴歸模型以分析師預測誤差、追蹤意願及預測離散程度三種特性分析財務分析師的預測行為，研究顯示相較於非家族企業，分析師對家族企業之預測意願較低，追蹤數量明顯較少，在預測結果上，家族企業會使分析師的預測產生較大的誤差，且各分析師間預測結果的差異程度也較大，本研究藉此結果推論家族企業中控制股東與其他股東代理問題的存在，及家族成員擔任管理者或董事等重要職位導致董事會喪失監督職能，在資訊揭露的數量與品質上表現較差。 / Prior research shows that family firms have grown rapidly and played an important role in the global corporate environment. This study examines the relation between family firms and financial analysts’ earnings forecast behaviors in Taiwan from year 2001 to 2008. I use several analysts forecast attributes: forecasts error, number of analysts following, and forecast dispersion. The results indicate that family firms generally have less analysts following, greater analysts’ forecast errors and greater forecast dispersion. These findings support the argument that the existence of conflict between majority and minority shareholders and that family members serving as managers or members of the board may weaken the disclosure of the quantity and quality of firm-specific information.

本研究旨在探討沙氏法302條款（Sarbanes-Oxley Act Section 302，以下簡稱SOX302）、沙氏法404條款（Sarbanes-Oxley Act Section 404，以下簡稱SOX404）與管理當局盈餘預測之關聯性，係美國S&P500之公司為研究對象，研究期間為2002年至2009年，主要檢測：(1)相對於SOX 404而言，在SOX 302規範下揭露無內部控制缺失之公司，其管理當局盈餘預測準確度是否會較低。(2)在SOX 404之規範下揭露有內部控制缺失之公司，公司層級的內部控制缺失相較於會計層級的內部控制缺失，是否會導致管理當局盈餘預測準確度較低。 本研究之實證結果顯示：(1)在無內部控制缺失情況下，相對於SOX 404而言，在SOX 302規範下，管理當局盈餘預測準確度會較低。(2)在SOX 404之規範下揭露有內部控制缺失之公司，公司層級的內部控制缺失相較於會計層級的內部控制缺失，會導致管理當局盈餘預測準確度較低。 / The main purpose of this thesis is to investigate the relation among SOX302, SOX404 and the accuracy of management earnings forecasts. Using a sample of S&P listed companies during the period 2002-2009, I address the following issues: (1) whether companies with effective internal controls under SOX302 have lower earnings forecast accuracy, compared with those under SOX404; (2) in the SOX404 context, whether companies with firm-level material weaknesses have lower earnings forecasts accuracy, compared with those with account-specific material weaknesses. As predicted, my empirical results show that: (1) compared with those under SOX404, companies with effective internal controls under SOX302 have lower earnings forecasts accuracy; (2) compared with those that disclosed account-specific material weaknesses under SOX404, the companies with firm-level material weaknesses have lower earnings forecasts accuracy.

本論文進行兩個眼動實驗，探討語境及多義特性在一詞多義處理歷程中扮演的角色。Frazier與Rayner(1990)認為一詞多義由於詞義間有重疊的語義屬性(sense overlap)，因此處理詞義之初，會先以部分語義(partial specification)來理解，不需立即進行完整詞義的解讀，支持部分語義解讀假設(Immediate Partial Interpretation Hypothesis)。在Frisson 與Pickering(2001)所提出詞義未定模型(Underspecification Model)中，進一步說明語境在一詞多義中扮演的角色，認為一詞多義的語義提取會先以未定意義(underspecified meaning)解釋，在語義提取完畢後，語境才會介入詞義的選擇。本論文操弄語境導引和詞義重疊性，來探討一詞多義的詞義提取與選擇歷程。實驗一操弄句子中目標詞的語境預測力(可預測語境vs.不可預測語境)與詞義數量(單詞義vs.多詞義)，記錄並分析目標詞與目標詞後區域的凝視時間。結果顯示：目標詞上的凝視時間多詞義顯著短於單詞義，支持部分語義解讀假設。然而，實驗一在目標詞後區域並未看到多詞義因延遲語義解讀(delayed semantic commitments)而造成抑制或競爭效果。實驗二就此現象與「一詞多義本身的詞義重疊程度」與「語境對一詞多義有無發生語義解讀的必要性」進行進一步探討。實驗二(同實驗一)操弄目標詞的語境預測力，與詞義的重疊程度(單義詞vs.中度詞義重疊vs.高度詞義重疊)，並控制語境偏向次要詞義的情形。結果顯示：目標詞上的晚期眼動指標在中度詞義重疊(中vs.單)與高度詞義重疊(高vs.單)皆看到與語境預測力發生交互作用，顯示語境預測力的影響在晚期階段發生；目標詞後區域觀察到中度與高度詞義重疊(中vs.單；高vs.單)皆在不可預測語境下發生詞義抑制效果，符合詞義未定模型預期之延遲語義解讀效果，並在目標詞晚期眼動指標看到中度詞義重疊效果與高度詞義重疊效果(單-中 vs. 單-高)受語境影響有不同的效果，顯示詞義重疊會影響詞義選擇的必要性。總結本論文的結果：首先，一詞多義的語義提取符合部分語義解讀的假設，而語境介入的影響支持詞義未定模型的看法，即語境在語義提取之初並未介入影響，在語義未定的情況下，若語境內容有進行語義解讀之必要時，則會發生延遲語義解讀。其次，詞義的促進效果與一詞多義的詞義數量多寡有關，詞義數量減少時，詞義促進效果也隨之消失。最後，當語境偏向一詞多義的次要詞義時，延遲語義效果才會發生。 / This thesis conducted two eye movement experiments with the aim to investigate the role of context and multi-sense feature in processing polysemy. Frazier and Rayner (1990) suggested that, at the beginning of semantic processing, polysemy is comprehended with partial specification. There is no immediate need to process the complete word sense due to the sense overlap between senses. This viewpoint supports the Immediate Partial Interpretation Hypothesis. In Frisson and Pickering’s (2001) Underspecification Model, further elaboration were made on the role of context in the process of retrieving and selecting one of the word senses of a polysemy. The sense that is first retrieved from a polysemy is considered to be an underspecified meaning. It is after the semantic retrieval is finished that context is involved in selecting a word sense. This thesis manipulated context guidance and sense overlap, to further research on processing polysemy in terms of word sense retrieval and selection. Experiment One manipulated contextual predictability (predictable context vs. unpredictable context) and number of senses (one-sense, monosemy vs. many-sense, polysemy) of the target words in sentences. Fixation times of the target words and post-target areas were recorded and analyzed. Results showed that the fixation times on target words were significantly shorter for polysemy than for monosemy, supporting the Immediate Partial Interpretation Hypothesis. However, in Experiment One, there is no inhibitory or competitive effect on the post-target area, indicating that there is no effect of delayed semantic commitments while comprehending polysemy. In Experiment Two, we further investigated how this phenomenon is connected with the degree of sense overlap and whether context is necessary to activate semantic commitment for polysemy. Contextual predictability of target words and the degree of sense overlap (monosemy vs. moderate-sense-overlap vs. high-sense-overlap) were manipulated, with the former designed as in Experiment One. Specifically, the context was controlled to bias toward the subordinate sense. The results showed that there were interactions of sense overlap degree (both moderate-sense-overlap vs. monosemy and high-sense-overlap vs. monosemy) and contextual predictability on target words for later-stage indices. This suggests that contextual predictability effects at later stages. On the post-target areas, there were inhibitory effects found for moderate-sense-overlap vs. monosemy and for high-sense-overlap when the context is unpredictable. This finding supports the delayed effect of semantic commitments in the Underspecification Model. Moreover, effects of sense overlap (polysemy-moderate vs. polysemy-high) were modulated by contextual predictability on target words for later-stage processing, showing that sense overlap affects the necessity of semantic commitments. In conclusion, the semantic retrieval of polysemy can be best explained by the Immediate Partial Interpretation Hypothesis and the involvement of contextual constraint supports the Underspecification Model. That is, context does not affect the beginning phase of semantic retrieval. Since the senses are underspecified, delayed semantic commitment occurs if it is necessary to make semantic commitment in the context. Furthermore, the facilitation of senses is related to the number of a polysemy’s senses. As number of senses decreases, facilitation of senses wanes and disappears. Finally, delayed semantic commitment occurs only when the context biases towards a subordinate sense of the polysemy.

台灣加權股價指數(TAIEX)，可以說是台灣最重要的經濟指數之一。在預測的方法中，時間序列分析一直都是熱門的課題，也是最常被使用來研究股價預測的方法。近年來，模糊理論在生醫、財務、社會、電機等各領域都有不錯的應用與發展 。本研究欲透過模糊區間的預測，主要是以時間序列預測台灣加權股價指數，來作為模糊區間精確度的探討，並針對區間時間序列進行模式的建構診斷和預測。最後我們將以2012年第一季(Q1)，每日交易股價指數的最高價與最低價作為實際研究的例子，同時也比較不同預測方法所得的結果。結果顯示模糊區間預測提供不同於傳統預測方法所得的資訊，希望能提供投資者另一種投資的參考。 關鍵字 : 台灣加權股價指數(TAIEX) 、模糊理論、模糊區間、區間預測 / Taiwan Weighted Stock Index (TAIEX) is one of Taiwan's most important economic indicators. Among the forecasting methods of time series analysis is always a hot issue on the forecasting methods and is also the most commonly used to make the stock price predictions. In recent years , fuzzy theory makes a great of application and development in various fields , such as , biomedical , financial and social …etc.. For this study, through the fuzzy interval forecasting is mainly based on time series forecasting TAIEX as fuzzy interval accuracy of the construction of diagnosis and prediction of the mode and interval time series. Finally, we will take the daily highest / lowest stock index prices data in the first quarter of 2012 (Q1) for actual research example , and will compare different forecasting methods of the results. The results show that the fuzzy interval forecasting differented from the traditional one on the basis of these information. We hope to offer investors an alternative investment advice. Keyword : Taiwan Capitalization Weighted Stock Index (TAIEX) 、 Fuzzy theory 、 Fuzzy interval、Interval forecasting.

財務預測相較於歷史性財務報表而言，其決策之攸關性較大。我國自民國80年5月起開始實施強制性財務報表迄今，已成為資本市場中相當重要的訊息，惟隨著我國證券市場規模逐步擴大，公司面臨之經濟環境日趨複雜，為符合國際作法及實務需要，故行政院金融監督管理委員會證券期貨局於93年12月9日依據證券交易法第三十六條之一，發布金管證六字第930005938號函，修正財務預測制度改採自願公開方式（並得以簡式或完整式方法擇一為之），並參考國外有關預測性資訊公開之相關規定及實務運作方式，配合相關配套措施，以增進資本市場效率及決策品質。 　　宣告取消財務預測強制公開揭露制度後，實證結果顯示：(1)對股價會產生一負向的效果，有顯著的負向異常報酬產生，且消息提前反應於市場上；在交易量方面，有顯著異常增加的情形發生。(2)對資訊特性不同分組採單變量分析，結果顯示公司規模及分析師人數可解釋宣告效果大小，兩組之間均達顯著水準。(3)公司規模越大時，股價反應越小，顯示公司規模與宣告異常報酬率間呈負向關係，亦即規模效果存在；但交易量反應越大。(4)分析師人數越多時，股價反應越小；但交易量反應越大，且較少分析師人數公司來得顯著。 / Financial forcast is more relevant than historical financial statements. Our country began to implement the mandatory financial forcast since May, 1991, has already become a very important information in the capital market. While the economic environment becoming complicated, in order to meet international practices, so on December 9, 1993, change the firm can dicide disclosure their financial forcast by themselves, in order to promote the efficiency of capital market and decision quality. The empirical results show that：(1) For stock price, it will produce negative effect; on the other hand, trading volume is significant increaseing. (2) For groups with different level of information characteristics, the results show that firm size and the number of ananlysts can explain that declaring effect. (3) When the size of firm is larger, the stock price is smaller; but the trading volume is larger. (4) The number of analysts is larger, the response of stock price is smaller; but the trading volume is larger in response.

我國的信用卡市場在民國八十二年全面開放以來，發展至今不過10餘年，已成為全球成長最快速的信用卡市場之一。但近年來也隨著信用卡業務已有相當顯著的成長，然而信用卡不僅只是一種支付工具，也屬於免擔保的信用融資，對發卡銀行而言，風險很高。故本文對於銀行要如何快速且正確的掌握客戶信用與還款能力，以防範呆帳發生，也變得日趨重要。 故本文利用Binary Quantile Regression可用於探討解釋變數對於被解釋變數在給定「特定分位數之下的邊際效果」，提供不同分位數的估計結果，可用於觀察被解釋變數的整個分配狀況。在實證上，二元分量迴歸模型不只可用來解釋平均的狀況，更常用來觀察分配尾端的情況。在以ROC與CAP的信用風險模型來驗證其Binary Quantile Regression的效力。

某些字彙可預告句子間的關係及文章架構。這些特殊的預測性字彙包含了連接詞及一部份的名詞、動詞、形容詞及副詞。本研究的目的是探討學習這類預測性字彙在提昇學生閱讀能力的成效。 研究對象是桃園縣某所高中104位三年級學生。研究主要工具包括一份評估學生閱讀能力的全民英檢考題；兩份問卷用來調查學生對於預測性字彙及文章架構的認知；及最後進行的訪談以了解教學成效。研究過程分為三個步驟: (1)前測--- 閱讀能力測驗及問卷。以測驗成績將分數高於平均的學生分至A組而分數低於平均的學生分至B組，來調查不同程度的學生對此教學的回應有無任何差異。而問卷則是調查學生對所要教授的預測性字彙及文章架構有無任何背景知識 (2)教學實驗---為期四個月的預測性字彙與文章結構教學 (3)後測--- 問卷及個別訪談，以瞭解學生在學過預測性字彙與文章結構後，在提昇閱讀能力方面有無任何成效。 研究結果發現，此教學實驗對學生的英文閱讀能力有正面影響。學生回應在閱讀時的速度及在興趣和信心這兩方面都比在接受教學之前提升。A組的學生在有關篇章結構的閱讀測驗題目中表現較B組學生好。同時A組的學生也比較知道如何使用預測性字彙來解析文章結構。至於學生常用的字彙為連接詞。 / Certain lexical items are strongly associated with identifiable clause and textual patterns. In other words, they represent the organization of discourse. These words include connectives and a set of open-system vocabulary with similar properties like connectives. The purpose of the study is to investigate how the instruction in lexical signals and pattern may facilitate reading comprehension. The researcher also examines what kind of predictive words are used most frequently in the process. One hundred and four students, from a senior high school in Taoyuan, were the participants of the study. The reading comprehension test of GEPT, intermediate level (LTTC, 2002), was used to classify the subjects into two sub-groups before the instruction. Group A was the one with the scores above the mean of the test scores. Group B was the one with the scores below the mean. The researcher classified the two groups so as to examine if the subjects in the two groups had any different feedback or response to the instruction. A questionnaire was applied to investigate if the subjects had any idea about the use of signals and patterns before the instruction. Then, a four-month signal and pattern instruction was conducted on all the participants in the study. After the instruction, the second questionnaire was applied to the subjects to gather their feedbacks and responses to the instruction. Furthermore, 30 subjects were randomly selected for interview to get in-depth information for qualitative analysis. The results show the instruction had a positive effect on the subjects when they read English texts. Their reading speed, interests and confidence increased after the teaching of organizing signals and text patterns. And they were confident in using signals to spell out clause patterns. However, some of the subjects had problems of identifying textual patterns, which usually involve several clause patterns combined together to form a text. The signals they used most frequently were connectives, the most overt signals in texts. And the subjects in Group A performed better than those in Group B when making use of predictive signals to spell out patterns. The result shows a strong relationship between the subjects’ English proficiency level and their manipulation of discourse-organizing words and textual patterns.

本論文主要在探討以移動平均法則為基礎的簡單技術分析指標，以及時間序列模型在台灣股票市場是否具有獲利能力，研究期間為1987/01/01-2006/12/31共20年的樣本期間。我們發現只有使用(1,50,0)和(1,50,0.01) 這兩個移動平均交易法則時才有顯著的報酬；並以AR(1)-GARCH(1,1)-M作為時間序列的預測模型。研究發現在股價上漲的時候，技術分析指標的確有較好的預測能力；而在股價下跌時，利用時間序列模型有較佳的獲利能力。因為技術分析指標與時間序列模型分別捕捉到不同的資訊，將兩預測工具結合在一起應該可以得到一個更好的組合預測指標。本文的實證研究發現此一組合預測指標，不管是在多頭或空頭期間時，都可以比使用單一分析工具獲得更高報酬。

本研究利用479個全球指標對台指期建立預測模型。該模型可以預測台指期在未來K天的漲跌走勢。我們使用了兩種演算法（支持向量機和決策樹）以及兩種取樣方式（交叉驗證和移動視窗）進行預測。在交叉驗證的建模過程中，決策樹展現了較高的預測力，最高準確度達到了93.4%。在移動視窗的建模過程中，支持向量機表現較好，達到了79.97%的預測准確度。於此同時，不管是哪一種條件設定都表明當我們預測的週期拉長時，預測的效果相對較好。這說明全球市場對台灣市場的影響很大，但是需要一定的市場反應時間。該研究結果對投資人有一定的參考作用。在未來方向裡，可以嘗試使用改進的決策樹演算法，也可以結合回歸預測進行深入研究。 / In this research, we build a stock price direction forecasting model with Taiwan Stock Index Future (TXF). The input data we used is 479 global indices. The classification algorithms we used are SVM and Decision Tree. This model can predict the up and down trend in the next k days. In the model building process, both cross validation and moving window are taking into account. As for the time period, both short term prediction (i.e. 1 day) and long term prediction (i.e. 100 days) are tested for comparison. The results showed that cross validation performs best with 93.4% in precision, and moving window reached 79.97% in precision when we use the last 60 days historical data to predict the up and down trend in the next 20 days. The results imply Taiwan stock market is significantly influenced by the global market in the long run. This finding could be further used by investors and also be studied with regression algorithms as a combination model to enhance its performance.

近年來，隨著大數據時代的來臨，即時資料運算面臨許多挑戰。例如在期貨交易預測方面，為了精準的預測市場狀態，我們需要在海量資料中建立預測模型，且耗時在數十毫秒之內。 在本研究中，我們將介紹一套即時巨量資料運算架構，這套架構將解決在實務上需要解決的三大需求：高速處理需求、巨量資料處理以及儲存需求。同時，在整個平行運算系統之下，我們也實作了數種人工智慧演算法，例如SVM (Support Vector Machine)和LR (Logistic Regression)等，做為策略模擬的子系統。本架構包含下列三種主要的雲端運算技術： 1. 使用Apache YARN以整合整體系統資源，使叢集資源運用更具效率。 2. 為滿足高速處理需求，本架構使用Apache Storm以便處理海量且即時之資料流。同時，借助該框架，可在數十毫秒之內，運算上千種市場狀態數值供模型建模之用。 3. 運用Apache Spark，本研究建立了一套分散式運算架構用於模型建模。藉由使用Spark RDD(Resilient Distributed Datasets)，本架構可將SVM和LR之模型建模時間縮短至數百毫秒之內。 為解決上述需求，本研究設計了一套n層分散式架構且整合上列數種技術。另外，在該架構中，我們使用Apache Kafka作為整體系統之訊息中介層，並支持系統內各子系統間之非同步訊息溝通。 / With the coming of the era of big data, the immediacy and the amount of data computation are facing with many challenges. For example, for Futures market forecasting, we need to accurately forecast the market state with the model built from large data (hundreds of GB to tens of TB) within tens of milliseconds. In this research, we will introduce a real-time big data computing architecture to resolve requests of high speed processing, the immense volume of data and the request of large data processing. In the meantime, several algorithms, such as SVM (Support Vector Machine, SVM) and LR (Logistic Regression, LR), are implemented as a subproject under the parallel distributed computing system. This architecture involves three main cloud computing techniques: 1. Use Apache YARN as a system of integrated resource management in order to apply cluster resources more efficiently. 2. To satisfy the requests of high speed processing, we apply Apache Storm in order to process large real-time data stream and compute thousands of numerical value within tens of milliseconds for following model building. 3. With Apache Spark, we establish a distributed computing architecture for model building. By using Spark RDD (Resilient Distributed Datasets, RDD), this architecture can shorten the execution time to within hundreds of milliseconds for SVM and LR model building. To resolve the requirements of the distributed system, we design an n-tier distributed architecture to integrate the foregoing several techniques. In this architecture, we use the Apache Kafka as the messaging middleware to support asynchronous message-based communication.