過去許多學術文獻針對租稅優惠吸引投資之效果進行實證分析，但實證結論並不一致。我國自1950年即開始實施一連串的租稅獎勵政策，時至今日，租稅減免仍是我國政府推動重大經濟政策慣用的誘因手段。為探討了解臺灣實施租稅優惠措施對投資變化之影響，本研究針對我國自民國50年代後期以來涉及租稅減免之相關法令進行整理，另為了充分量化這些租稅優惠措施，則參考國外相關實證文獻作法，建構了兩項租稅誘因指標作為虛擬變數，來追蹤自民國61年以來我國不同階段之減免稅狀態，並作為虛擬變數納入後續實證模型分析。 本研究利用相關變數之時間序列資料來探討租稅優惠對我國外人直接投資與國內私人投資之互動變化關係。研究步驟有三，首先，針對個別變數進行單根檢定，藉以確認變數的屬性，了解時間序列資料是否為衡定後，再利用Johansen共整合檢定法來估計和檢定多個變數，確認各變數間是否存在共整合關係後，以誤差修正模型來說明各變數間關係與整個變數脫離均衡關係後之動態調整情形。 實證結果顯示，就長期趨勢而言，我國實施之租稅優惠措施對吸引外人直接投資呈現負向且顯著之不良影響，另長期而言，租稅優惠誘因對刺激我國國內投資之變化確有顯著且正面助益，但影響效果之幅度不大。因此，本研究建議政府與其提供效果不明確之租稅誘因，不如致力於針對國家自身不完善的基礎建設或不穩定的總體經濟環境進行改善。 / Tax incentives have been in existence in Taiwan since 1950, and they are still very much on the agenda of the government. There is no agreement about the efficacy of incentives. Indeed there have been doubts about whether incentives have any effect on the economy since the 1950s. This has made some economists wonder why incentives are so popular despite the fact that their effects are either slight or unknow. This study conducts an empirical investigation of the impact of tax incentives on investment in Taiwan. We constructed two indexs of tax incentives which track the different types of incentives embarked upon by the government, and these indexes are then included in both foreign direct investment and private investment equations. Our testing procedure involves three steps. The first step involves tasting for the properties of the variables by conducting unit root teste. The second step involves testing for the long-run relationship between the variables using Johansen cointegration tests. And the third step involves estimating the long-run parameters and associated loading factors. The empirical results shows a significant negative impact of tax incentives on FDI, and a significant positive impact of tax incentives on private investment but the impact is slight. We suggest that rather than focusing on tax incentives, the country should concentrate on removing the factors that discourage investors such as poor infrastructural and institutions or macroeconomic instability.

因網際網路帶來的便利性與即時性，網路新聞成為社會大眾吸收與傳遞新聞資訊的重要管道之一，而累積的巨量新聞亦可反映出社會輿論對某特定新聞議題之即時反應、熱門程度以及情緒走向等。 因此，本研究期望借由意見探勘與情緒分析技術，從特定領域新聞中挖掘出有價值的關聯，並結合傳統機器學習建立一個房地產市場的預測模式，提供購屋決策的參考依據。 本研究搜集99年1月1日至103年6月30日共1,1150筆房地產新聞，以及8,165件捷運週邊250公尺內房屋買賣交易資料，運用意見探勘萃取意見詞彙進行情緒分析，並建立房市情緒與成交價量時間序列，透過半年移動平均、二次移動平均及成長斜率，瞭解社會輿情對房市行情抱持樂觀或悲觀，分析社會情緒與實際房地產成交間關聯性，以期能找出房地產買賣時機點，並進一步結合情緒及房地產的環境影響因素，藉由支援向量機建立站點房市的預測模型。 實證結果中，本研究發現房市情緒與成交價量之波動有一定的週期與相關性，且新捷運開通前一年將連帶影響整體捷運房市波動，當成交線穿越情緒線且斜率同時向上時，可做為適當的房市進場時機點。而本研究針對站點情緒與環境變數所建立之預測模型，其預測新捷運線站點之平均準確率為69.2％，而預測新捷運線熱門站點之準確率為78％，顯示模型於預測熱門站點上具有不錯的預測能力。 / Nowadays, E-News have become an important way for people to get daily information. These enormous amounts of news could reflect public opinions on a particular attention or sentiment trends in news topics. Therefore, how to use opinion mining and sentiment analysis technology to dig out valuable information from particular news becomes the latest issue. In this study, we collected 1,1150 house news and 8,165 house transaction records around the MRT stations within 250 meters over the last five years. We extracted the emotion words from the news by manipulating opinion mining. Furthermore, we built moving average lines and the slope of the moving average in order to explore the relationship and entry point between public opinion and housing market. In conclusion, we indicated that there is a high correlation between the news sentiment and housing market. We also uses SVM algorithm to construct a model to predict housing hotspots. The results demonstrate that the SVM model reaches average accuracy at 69.2% and the model accuracy increases up to 78% for predicting housing hotspots. Besides, we also provide investors with a basis of entry point into the housing market by utilizing the moving average cross overs and slopes analysis and a better way of predicting housing hotspots.

我國健保費用逐漸增長，進而衍生出許多健保問題，其中浮報、虛報及詐欺等三種情況，會造成許多醫療資源的浪費。然而，目前電腦檔案分析只能偵測出浮報、虛報的行為，無法偵測出詐欺情況。對於健保詐欺之偵測只能仰賴傳統隨機抽樣檢驗及人力分析，而我國健保平均一年門診審查申報量約3.5 億件，其人力的負擔非常沉重。故本研究將探討如何利用電腦工具初步判別醫事機構之費用申報情況。 本研究透過大量文獻回顧，發現美國有研究指出結合Benford’s law 與智慧型方法來進行詐欺偵測，可獲得很好的效果(Busta & Weinberg 1998)。Benford’s law 指出許多數據來源皆會呈現特定的數字頻率分佈，近年來Benford’s law 亦被應用在許多不同領域的舞弊或詐欺的審查流程中。 本研究使用Apache Hadoop 及其相關專案，建構出一個大量資料儲存分析之環境，針對大量健保申報費用資料來進行分析。此系統結合了Benford’s law 數字分析方法並運用支持向量機(Support Vector Machine)來對健保費用申報進行大規模電腦初步審查，判別該醫事機構是否有異常申報之情況發生，並將初步判別之結果提供給健保局相關稽查人員，進而做深入的審查。 本研究所建構的智慧型健保費用異常偵測模型結合了Benford’s law 衍生指標變數與實務指標變數，並利用SVM 分析健保申報費用歷史資料，產生出預判模型，之後便可藉由此模型來判別未來健保費用申報資料是否有異常情況發生。在判別異常資料方面，本研究所建構的模型其整體正確率高達97.7995%，且所有的異常申報資料皆可準確地預測出來。 因此，本研究希望能結合Benford’s law 與智慧型運算方法於健保申報異常偵測上，如此一來便可藉由電腦進行初步審查，減少因傳統隨機抽樣調查所造成的不確定性以及審核大量健保資料時過多的人力資源浪費。

全球智慧型手機的出貨量持續提升，且熱門市集的App下載次數紛紛突破500億次。而在iOS和Android手機App市集中，App的評價和評論對App在市集的排序有很大的影響；對於App開發者而言，透過評論確實可掌握使用者的需求，並在產生抱怨前能快速反應避免危機。然而，每日多達上百篇的評論，透過人力逐篇查看，不止耗費時間，更無法整合性的瞭解使用者的需求與問題。 文字情感分析通常會使用監督式或非監督式的方法分析文字評論，其中監督式方法被證實透過簡單的文件量化方法就可達到很高的正確率。但監督式方法有無法預期未知趨勢的限制，且需要進行耗費人力的文章類別標注工作。 本研究透過情感傾向和熱門關注議題兩個面向來分析App評論，提出一個混合非監督式與監督式的中文情感分析方法。我們先透過非監督式方法標注評論類別，並作視覺化整理呈現，最後再用監督式方法建立分類模型，並驗證其效果。 在實驗結果中，利用中文詞彙網路所建立的情感詞集，確實可用來判斷評論的正反情緒，唯判斷負面評論效果不佳需作改善。在議題擷取方面，嘗試使用兩種不同分群方法，其中使用NPMI衡量字詞間關係強度，再配合社群網路分析的Concor方法結果有不錯的成效。最後在使用監督式學習的分類結果中，情感傾向的分類正確率達到87%，關注議題的分類正確率達到96%，皆有不錯表現。 本研究利用中文詞彙網路與社會網路分析，來發展一個非監督式的中文類別判斷方法，並建立一個中文情感分析的範例。另外透過建立全面性的視覺化報告來瞭解使用者的正反回饋意見，並可透過分類模型來掌握新評論的內容，以提供App開發者在市場上之競爭智慧。 / While the number of smartphone shipment is continuesly growing, the number of App downloads from the popular app markets has been already over 50 billion. By Apple App Store and Google Play, ratings and reviews play a more important role in influencing app difusion. While app developers can realize users’ needs by app reviews, more than thousands of reviews produced by user everday become difficult to be read and collated. Sentiment Analysis researchs encompass supervised and unsupervised methods for analyzing review text. The supervised learning is proven as a useful method and can reach high accuracy, but there are limits where future trend can not be recognized and the labels of individual classes must be made manually. We concentrate on two issues, viz Sentiment Orientation and Popular Topic, to propose a Chinese Sentiment Analysis method which combines supervised and unsupervised learning. At First, we use unsupervised learning to label every review articles and produce visualized reports. Secondly, we employee supervised learning to build classification model and verify the result. In the experiment, the Chinese WordNet is used to build sentiment lexicon to determin review’s sentiment orientation, but the result shows it is weak to find out negative review opinions. In the Topic Extraction phase, we apply two clustering methods to extract Popular Topic classes and its result is excellent by using of NPMI Model with Social Network Analysis Method i.e. Concor. In the supervised learning phase, the accuracy of Sentiment Orientation class is 87% and the accuracy of Popular Topic class is 96%. In this research, we conduct an exemplification of the unsupervised method by means of Chinese WorkNet and Social Network Analysis to determin the review classes. Also, we build a comprehensive visualized report to realize users’ feedbacks and utilize classification to explore new comments. Last but not least, with Chinese Sentiment Analysis of this research, and the competitive intelligence in App market can be provided to the App develops.

本文利用向量自我迴歸模型所得出來的殘差值來模擬未預期到的總體經濟訊息，以期限利差和一個月定存利率來捕捉殖利率曲線，以違約利差和股利收益率來描繪資產報酬的條件機率分布，本文實證未預期到的期限利差和未預期到的違約風險與淨值市價比因子和市值規模因子包含相同的訊息，因此後續檢驗這些能夠捕捉未來投資機會的總體經濟訊息比起Fama-French三因子模型是否對台灣股票橫斷面的平均報酬更具有解釋能力。 實證方法採用Fama-MacBeth(1973)兩階段迴歸方法，Fama-French三因子模型實證結果顯示台灣股票市場存在著負向的淨值市價比效果，但卻不存在著規模效果，這與國外一些學者研究1980年代之後規模效果逐漸消失的結論相同。在實證未預期到的總體經濟訊息模型時，由於被解釋變數為股票超額報酬率，因此常數項應該為不顯著的關係，但此假設強烈的被未預期到的總體經濟訊息模型拒絕，代表此模型可能遺漏了重要的解釋變數。因此，Fama-French 三因子模型對台灣股票橫斷面平均報酬率的解釋能力比未預期到的總體經濟訊息模型更佳。 / The Fama and French factors HML and SMB are correlated with innovations in variables that describe investment opportunities. I find that shocks to term spread and shocks to default spread have the same information with the Fama and French factors HML and SMB. This paper investigates whether a model that includes shocks to the aggregate dividend yield and term spread, default spread, and one-month deposit interest rate can explain the cross section of average return on Taiwan stock market as well as the Fama and French can. Using the Fama-MacBeth (1973) two steps cross-sectional regressions, I find there exists the negative book-to-market effect on Taiwan stock market, but the size effect disappears. Since the dependent variables in the regression is excess returns, the intercept of the cross-sectional regression should be zero. This hypothesis is strongly rejected in the case of the model includes shocks to the Macroeconomics variables and the market portfolio. It means this model omits some important variables, so the Fama and French three-factor model can explain the cross section of average returns better.

本文使用Pesaran,Schuermann and Weiner(2004)提出的全球化向量自我迴歸模型(Global Vector Autoregression Model, GVAR)對房地產市場進行分析。 我們考慮 1994Q1 至 2011Q2 的資料，納入美國、中國、日本及台灣，每個國家各七個變數及一個國際外生變數，使用衝擊反應函數去衡量總體經濟變數與房市之間的連動性，以及房地產市場在國際之間的外溢效果。 本文針對美國實質房價衝擊、美國實質產出衝擊及台灣實質產出衝擊做探討，所得到的實證結果主要可歸納為三點:首先是美國房市下跌會傳遞至其它經濟面，如實質產出、通膨率、利率市場，影響會在第四季時恢復平穩，但多存在著長期影響。其二為當美國景氣衰退時，美國利率市場的反應較為迅速，中國、日本及台灣平均會落後一到兩季才會反應，且美國利率的反應幅度會較大。最後一點為跨國之間的房地產市場雖然沒有顯著的直接連動關係，但是會透過不同管道間接影響他國的經濟市場，其中一個管道可能是經由財富效果傳導至實質經濟面，造成消費需求上的衝擊，進而影響兩國的貿易平衡，另一方面則可能會影響各國央行的貨幣政策，透過金融管道對跨國間的投資產生影響。 關鍵詞:全球化向量自我迴歸模型、共整合、誤差修正模型、房地產市場、財富效果。

由於採用的地球原子不同，目前，台灣地區有兩種坐標系統存在，TWD67(Taiwan Datum 1967) 和TWD97(Taiwan Datum 1997)。在應用上，必須進行不同地球原子間之坐標轉換。坐標轉換方面，有許多方法可供選擇，如六參數轉換、支持向量機(Support Vector Machine, SVM)轉換等。 最小二乘支持向量機(Least Square Support Vector Machine, LSSVM)，為SVM的一種演算法，是一種非線性模型。LSSVM在運用上所需之參數少，能夠解決小樣本、非線性、高維度和局部極小點等問題。目前，LSSVM，已經被成功運用在影像分類和統計迴歸等領域上。 本研究將利用LSSVM採用不同之核函數：線性核函數(LIN)、多項式核函數(POLY)及徑向基核函數(RBF)進行TWD97和TWD67之坐標轉換。研究中並使用基因演算法來調整LSSVM的RBF核函數之系統參數(後略稱RBF+GA)，找出較佳之系統參數組合以進行坐標轉換。模擬與實測之地籍資料，將被用以測試LSSVM及六參數坐標轉換方法的轉換精度。 研究結果顯示，RBF+GA在各實驗區之轉換精度優於參數優化前RBF之轉換精度，且RBF+GA之轉換精度也較六參數轉換之轉換精度高。 進行參數優化後，RBF+GA相對於RBF的精度提升率如下：(1)模擬實驗區：參考點與檢核點數量比分別為1：1、2：1、3：1、1：2及1：3時，精度提升率分別為15.2%、21.9%、33.2%、12.0%、11.7%；(2)真實實驗區：花蓮縣、台中市及台北市實驗區之精度提升率分別為20.1%、32.4% 、22.5%。 / There are two coordinate systems with different geodetic datum in Taiwan region, i.e., TWD67 (Taiwan Datum 1967) and TWD97 (Taiwan Datum 1997). In order to maintain the consistency of cadastral coordinates, it is necessary to transform from one coordinate system to another. There are many coordinate transformation methods, such as, 2-dimension 6-parameter transformation, and support vector machine (SVM). Least Square Support Vector Machine (LSSVM), is one type of SVM algorithms, and it is also a non-linear model。LSSVM needs a few parameters to solve non-linear, high-dimension problems, and it has been successfully applied to the fields of image classification, and statistical regression. The goal of this paper is to apply LSSVM with different kernel functions (POLY、LIN、RBF) to cadastral coordinate transformation between TWD67 and TWD97. Genetic Algorithm will be used to find out an appropriate set of system parameters for LSSVM with RBF kernel to transform the cadastral coordinates. The simulated and real data sets will be used to test the performances, and coordinate transformation accuracies of LSSVM with different kernel functions and 6-parameter transformation. According to the test results, it is found that after optimizing the RBF parameters (RBF+GA), the transformation accuracies using RBF+GA are better than RBF, and even better than those of 6-parameter transformation. Comparing with the transformation accuracies using RBF, the transformation accuracy improving rate of RBF+GA are : (1) The simulated data sets: when the amount ratio of reference points and check points comes to 1:1, 2:1, 3:1, 1:2 and 1:3, the transformation accuracy improving rate are 15.2%, 21.9%, 33.2%, 12.0% and 11.7%, respectively; (2) The real data sets： the transformation accuracy improving rate of RBF+GA for the Hualien, Taichung and Taipei data sets are 20.1%, 32.4% and 22.5%, respectively.

傳統的腫瘤指標篩檢方法，往往靈敏度、普及度及特異性有限，無法得到正確、即時的診斷結果。現今癌症的研究，則透過蛋白質體學經由光譜及影像觀察癌症不同時期的蛋白質表現變化，期望未來得以發展較佳之診斷工具。本研究中主要針對兩組攝護腺癌症病人之蛋白質質譜資料，此資料應用蛋白質晶片與表面強化雷射解吸電離飛行質譜技術（SELDI-TOF-MS）收集而來。我們的研究目的在於從大量的蛋白質特徵中篩選出一群有助於分類的蛋白質特徵變數。我們提出以最小分錯率特徵選取法與最小p值（ 檢定、Kruskal-Wallis檢定）特徵選取法進行初步特徵辨識度排序以及選取，並進一步發展出k-mean萃取法、最大相關係數萃取法與判定係數萃取法以改善變數間嚴重的共線性問題。我們利用支援向量機（Support Vector Machine）方法進行分類並評估分類效果，在不同的分類目的下萃取有助於辨識的蛋白質特徵，以決定最佳特徵集合。研究發現運用最小分錯率特徵選取法與最小p值分錯率特徵選取法，輔以判定係數萃取法，在各分類目的下皆有良好表現，為較佳的特徵選取方式。

銅價在近年來不斷的創下歷史新高，由於台灣蓬勃的電子、半導體、工具機產業皆需要銅，因此銅進口量位居全球第五(ICSG,2009)，使得台灣企業的生產成本受國際銅價的波動影響甚鉅，全球有70%的銅價是按照英國倫敦金屬交易所(London Metal Exchange, LME)的牌價進行貿易，因此本研究欲建置預測模式以預測銅價未來趨勢。 　　本研究之資料來源為2003年1月2日至2011年7月14日的LME三月期銅價，並依文獻探討選取LME的銅庫存、三月期鋁價、三月期鉛價、三月期鎳價、三月期鋅價、三月期錫價，以及金價、銀價、石油價格、美國生產者物價指數、美國消費者物價指數、聯邦資金利率作為影響因素的分析資料。時間序列分析、類神經網路已被廣泛的用於預測股市及期貨，本研究先藉由向量自我迴歸模型篩選出有影響力的變數，同時建置GARCH時間序列預測模型與具有遞迴的Elman類神經網路預測模型，再整合兩者建置GARCH-Elman類神經網路預測模型。 本研究之向量自我迴歸模型顯示銅價與金、鋁、銅庫存前第1期；自身前第2期；鎳、錫前第3期；鋅前第4期的變動有負向的影響；受到石油前第2期的變動有正向的影響，這其中以銅的自我解釋變異最高，銅庫存最低，推測其影響已有效率地反映到銅價上。也驗證預測模型必須考量總體經濟變數，且變數先經向量自我迴歸模型的篩選能因減少雜訊而提升類神經網路的預測能力。依此建置的GARCH模型有33.81%的累積報酬率、Elman類神經網路38.11%、整合兩者的GARCH-Elman類神經網路56.46%，皆優於實際銅價指數的累積報酬率。對銅有需求的企業者，能更為準確的預測漲跌趨勢，依此判斷如何跟原物料供應商簽訂合約的價格與期間，使其免於價格趨勢的誤判而提高生產成本，並提出五點建議供未來研究者參考。 / The recent copper price in London Metal Exchange (LME) has breaking the historical high. Taiwan’s booming electronics, semiconductor and machine tool industry causing copper import volume ranked fifth in the world (ICSG, 2009). Because of 70% of copper worldwide trade in accordance with the price of the London Metal Exchange, this study using time series and neural networks to build the LME copper price forecast model. 　　This study considering copper, copper stocks, aluminum, lead, nickel, zinc, tin, gold, silver, oil ,federal funds rate, CPI and PPI during the period of 2003/1/2 to 2011/7/14. Time series model and neural networks have been widely used for forecasting the stock market and futures. In this study, using Vector Autoregressive (VAR) model screened influential variables, building GARCH model and Elman neural network to forecast the LME copper price; and further, integrating this two models to build GARCH-Elman neural network prediction model. 　　This study’s VAR models show that the copper has negative effect with gold, aluminum, copper stocks, nickel, tin, zinc and itself. And has positive impact with oil prices. The highest of explained variance is copper. Copper stocks are lowest, speculating that its impact has been efficiently reflecting on the price of copper. Verifying the prediction model must consider the macroeconomics variables. Using VAR model screened influential variables can reduce noise to enhance the predictive ability of the neural network. This study’s GARCH model has 33.81% of the cumulative rate of return, Elman neural network has 38.11% and the GARCH-Elman neural network has 56.46%. All of them are better than the actual price of copper.

台灣不動產投資信託（T-REITs）自2005年發行至今已逾六年，然其市場表現仍不如發行之初所預期。過去國內已有許多研究針對T-REITs市場發展進行探討，然而目前就T-REITs與直接不動產投資市場價格表現間之相關研究尚付之闕如。有鑑於此，本研究藉由共整合與Granger因果關係檢定，檢視REITs與直接不動產市場間之關聯性，了解台灣與美國之REITs市場表現差異及其影響因素，進而作為改進T-REITs運作機制或架構之參考依據。 實證結果發現，美國之REITs與直接不動產市場之間存在共整合關係。此結果表示，長期而言，這兩者可能具有相似之風險分散效益。此外，透過Granger因果關係檢定發現REITs領先於直接不動產，乃因前者市場較具效率。另一方面，台灣之REITs與直接不動產市場之間則不具有共整合以及領先或落後關係，然直接不動產當期價格仍會受到本身與REITs之前期價格影響。 本研究進一步分析台、美兩國實證結果之差異原因如下：資料的樣本期間、REITs市場規模、存在於T-REITs市場之集中性風險以及潛在的代理問題。其中，針對T-REITs潛在代理問題，本研究藉由分析股票與T-REIT報酬率之波動性，發現T-REIT之不動產管理機構若與母集團相關者，則其市場表現較差。因此，我們得出T-REITs市場發展主要是受限於代理問題之結論。本研究成果不僅有助於改善T-REITs市場效率，亦可提供學術與實務之參考。 / The mechanism of Real Estate Investment Trusts in Taiwan (or T-REITs) was launched in 2005, however, T-REITs market did not perform as expected. What caused the limited development of T-REITs market? Current literature on the performance between T-REITs and direct real estate investment is limited. Through the cointegration and Granger causality tests, the purpose of this study is hence to explore the short-term and long-term dynamics between REITs and direct real estate markets in the U.S. and Taiwan, respectively. This study presents evidence of the cointegration relationship between REITs and direct real estate in the U.S. It implies that the diversification properties of these two assets are likely to be similar over the long horizon. According to the Granger causality test, REITs leads direct real estate due to the market information efficiency. These findings are consistent with those of previous studies. On the other hand, we find no cointegration and lead-lag relation between T-REITs and commercial real estate. Moreover, the current commercial transaction price is affected by both its and T-REIT previous price. By comparing the difference between the results of these two countries, there are several possible explanations for the different results between the U.S. and Taiwan, including difference in sample period, market capitalization, concentrated risk, and most importantly, the potential agency problem existing in T-REITs market. Finally, the underperformance of parent-related management T-REIT is verified through the volatilities of stock and T-REIT returns. Therefore, we conclude that the limited development of T-REITs is caused by the agency problem in REITs market. Results of this study may provide T-REITs market for improving its efficiency, as well as for the reference for both academics and real practices.