国立情報学研究所で電子化したコンテンツを使用している。

国立情報学研究所で電子化したコンテンツを使用している。

開放政府資料 (Open Government Data)現已成各國治理趨勢，政府盼藉開放所持資料，允許民間加值應用後，能夠促進公民參與、刺激創新發展、改善社會問題。聯合國2016年電子化政府報告指出透過政府開放資料可望實現永續發展指標，其中環境議題的永續發展為重點目標之一。我國政府開放資料表現亮眼，於2015年獲得全球開放資料普查(2015 Global Open Data Index)第一名的佳績，惟與世界各國的開放資料政策面臨相同的「缺乏使用者」的問題。因此本研究欲瞭解我國環保團體作為環境議題的主要倡議者，其應用環境開放資料推動永續發展情形。 本研究採半結構式訪談法，以我國有/無使用環境開放資料的環保團體及資料處理專家為對象進行訪談，瞭解其使用或不使用開放資料推動永續發展訴求之因素。透過本研究訪談成果發現，主要因素有四：一為使用動機，若該環團訴求為改善整體環境政策，亟需精準環境資料佐證，才有動機使用開放資料，若其訴求為個案議題即沒有動機使用；二為環團內部數據分析人才，若環團內部有基本數據分析人才，將有助環團瞭解開放資料對其環境訴求助益，有利於後續推廣應用；三為環團對政府開放的資料正確性仍有不信任感，即便使用亦是抱持交叉查證心態使用；四為政府開放的環境資料專業門檻高，環團應有一定的環境專業才能夠解讀及應用開放資料。

本研究以台灣50與中型100成分股為對象，運用資料探勘特徵選取技術，分析影響股票完全填權息成功之關鍵因素，並依此關鍵因素建構一個完全填權息預測模型，最後比較研究結果與過去研究之異同。本研究完全填權息預測模型的建構過程分為五階段：(1)定義完全填權息之股票：運用TEJ資料庫抓到的歷史股價資料與股利資訊，計算除權息前與除權息後股價，標註完全填權息和未完全填權息二個類別。(2)影響填權息相關因素：根據過去文獻所發現，影響短期填權息行情超額報酬的因素，以及影響股價的基本面因素，蒐集與股利相關的指標與基本分析中所用的公開財務報表資料。(3)特徵選取分析：利用循序前進搜尋（SFS）結合分類演算法，整合與計算所有影響因素資料，藉此找出關鍵的影響因素。(4)預測模型建立：根據特徵選取之結果資料，使用Weka軟體進行資料探勘支持向量機和決策樹分類模型訓練。(5)模型準確性比較與分析：本研究所建構之模型可協助存股型投資者，判斷可領取高股息且無股價損失之股票，提供投資人選股參考。 / In this study, we use the Feature Selection Method for Data Mining to analyze the key factors that may affect the rate of the stock price successfully remaining previous price after dividend payment among stocks of 50 largest companies and 100 medium-sized companies in Taiwan. Based on these key factors, we construct a forecasting model for stocks with the 100% flat stock price. Finally, We try to find out the similarities and differences between the current study and past research. In this study, the construction of a forecasting model for stocks with the 100% flat stock price is divided into five stages: (1) Defining stocks with the 100% flat stock price: Marking stocks with the 100% flat stock price and the non-100% flat stock price on historical stock data and dividend information captured by the TEJ database; (2) Relevant Factors Affecting increase in the stock price after dividend payment: According to the factors found in the past literature that may affect excess returns from short-term increase in the stock price after dividend payment and the fundamental factors affecting the stock price, we are able to collect indexes related to dividends and public financial statements for basic analysis. (3) Feature Selection Analysis: By using the Sequential Forward Selection (SFS) method and the classification algorithm, all influencing factors are integrated and calculated to find out the key influencing factors; (4) The Establishment of the Prediction Model: According to the results of feature selection, we use the Weka software to conduct data mining and train the classification model based on support vector machines and decision trees. (5) Comparison and Analysis on Accuracy of the Model: The model constructed in this study can help stock-holding investors determine stocks with high dividends without loss of the stock price and provide reference for investors in stock selection.

我國於2015年底通過新修正之個人資料保護法（以下簡稱「個資法」），將病歷納入特種個人資料中保護。目前個資法第六條特種個人資料列舉包含病歷、醫療、基因、性生活、健康檢查及犯罪前科之個人資料。雖然該條文係取法自國際賦予敏感性個人資料特別保護的模式，惟在個人相關健康資料保護部分，我國個資法不若歐盟一般資料保護規則（EU General Data Protection Regulation, GDPR）保護寬廣，納入資料之類型仍較國際立法例狹窄。尤其此次GDPR修法擴大特種個人資料空間，增列基因資料、生物性資料和性傾向，檢視我國特種個人資料列舉類型是否符合現今科技社會需求有其必要性。 過去研究針對健康資料個資法適用問題較少。大數據資料來源來自各處，以一般健康保健物聯網模式為例，自行操作之檢查數據或穿戴式裝置所蒐集之資料，若非須由醫師或其他之醫事人員施以檢查，而可由一般民眾自行測量之行為，該民眾自行測量之結果應不屬於個資法所謂之病歷、醫療或健康檢查個人資料，即非為特種個人資料。 惟大數據分析技術進步之環境下，健康資料亦攸關資料主體生理健康之敏感性，且容易連結並識別個人，考量健康資料敏感性提升，蒐集、處理、利用健康資料易侵犯到個人隱私，因此有加強保護之需求。將來可刪除個資法第六條第一項各種個人資料例示之「醫療」、「病歷」與「健康」資料，並新增「健康」或「與健康相關」之列舉項目。 但解釋「與健康相關」資料之內涵時不能無限上綱，在適用時應考量情境說，依據不同使用情境判斷是否為係作為特種個人資料利用，以排除一般性描述健康的使用情境。 / The change to the regulation of special categories of data (sensitive data) in the Taiwan Personal Information Protection Act (PIPA) in 2015 comes with the inclusion of medical records. The definition of sensitive data in the PIPA Article 6(1) refers to personal information of medical records, medical treatment, genetic information, sexual life, health examination and criminal records. However, the list of sensitive data in PIPA do not contain categories as broad as foreign legislation such as EU General Data Protection Regulation (GDPR). It is important to review the continuing relevance of existing categories of sensitive data in the light of change in social structures and advances in technology. Differ from “medical data” such as medical records, medical treatment and health examination, the collection, process and use of “health data” which is measured from wearable device, is not included in the sensitive data. Concerning the development of big data analysis, the “health data” which sensitivity enhanced is easy to identify an individual. It needs to give a higher level of protection to “health data” under PIPA. Therefore, this thesis suggests that medical records, medical treatment and health examination in PIPA Article 6(1) should be consolidated and amended to health records or data concerning health. However, this is not to say that the processing of all kinds of medical and health data should be regarded as the processing of sensitive data. But data, under certain contexts/circumstances may be treated as the processing of sensitive data.

此次的研究要是要探討人口老化是否會直接或間接影響儲蓄率升降，並且針對近18年來台灣20各縣市儲蓄率的變動來進行探討，利用文獻探討中所提及的各項解釋變數，再加入此次的重點人口老化這個變數來檢驗分析人口老化是否會影響儲蓄率的變動。

近年來鏈結開放式資料 (Linked Open Data，簡稱LOD) 被認定含有大量潛在價值。如何蒐集與整合多元化的LOD並提供給資料分析人員進行資料的萃取與分析，已成為當前研究的重要挑戰。LOD資料是RDF (Resource Description Framework) 的資料格式。我們可以利用SPARQL來查詢RDF資料，但是目前對於大量RDF的資料除了缺少一個高性能且易擴展的儲存和查詢分析整合性系統之外，對於RDF大數據資料分析流程的研究也不夠完備。本研究以預測電影票房為例，使用DBpedia LOD資料集並連結外部電影資料庫 (例如：IMDb)，並在Spark大數據平台上進行巨量圖形的分析。首先利用簡單貝氏分類與貝氏網路兩種演算法進行電影票房預測模型實例的建構，並使用貝氏訊息準則 (Bayesian Information Criterion，簡稱BIC) 找到最佳的貝氏網路結構。接著計算多元分類的ROC曲線與AUC值來評估本案例預測模型的準確率。 / Recent years, Linked Open Data (LOD) has been identified as containing large amount of potential value. How to collect and integrate multiple LOD contents for effective analytics has become a research challenge. LOD is represented as a Resource Description Framework (RDF) format, which can be queried through SPARQL language. But large amount of RDF data is lack of a high performance and scalable storage analysis system. Moreover, big RDF data analytics pipeline is far from perfect. The purpose of this study is to exploit the above research issue. A movie box office sale prediction scenario is demonstrated by using DBpedia with external IMDb movie database. We perform the DBpedia big graph analytics on the Apache Spark platform. The movie box office prediction for optimal model selection is first evaluated by BIC. Then, Naïve Bayes and Bayesian Network optimal model’s ROC and AUC values are obtained to justify our approach.

顧客關係管理(CRM)在國內外已有不少應用實例，但在醫療服務業鮮少被研究過，本研究嚐試將資料採礦的三大核心技術：資料庫管理、Domain知識與資料採礦技術三者予以整合，針對一個獨立經營主體（聯合診所），從行銷策略制定、營運策略描述與執行到經由資料採礦得到具體結果，重新檢討行銷策略之STP定位與導引未來經營策略，並提出一對一行銷的診所病患管理架構。 本研究以一個新成立的診所，取其開業之初(89年12月)至92年1月底止，所累積九千三百多位患者的5萬多筆門診就醫記錄進行資料採礦分析希望研究以下幾個問題： 1.哪些病患帶來最大利潤?為甚麼?哪些患者容易流失?為甚麼? 2.哪些交叉服務對何種患者適合?哪些服務對增加慢性病患者有幫助?糖尿病患者接受視網膜檢查的可能原因為何?婦產科門診所增加的病患,是否會同時接受診所內其他科的服務?是否應該繼續擴大其他專科? 3.診所病患主要的居住地區如何描述？ 研究結果顯示較高獲利組與高醫療費用,高忠誠度,高就診次數,高藥費比率,高慢性病費用比率有關，以疾病別來看，集中在慢性疾病患者身上。顯然經營策略上的意涵是如何爭取慢性病人的高度滿意及信賴度，贏得高忠程度，患者願意將診所視做健康上的守門人(Gate Keeper)，而從地區別分析中也發現一些，診所服務之涵蓋範圍，可以區分為距離效益、慢性病患者口碑效應與轉移效益。慢性病患者之分群可以分成黃金老主顧、會忘記看病的老主顧、快流失的老主顧、高穩定低忠誠度高獲利新客戶、不常來但還會來的一般客戶、已流失的舊客戶、已流失的中期客戶及流失已久的舊客戶，至於非慢性病患則不需太複雜的分群，本研究建議將非慢性病患者依健保卡卡序計算就醫忠誠度區分。慢性病患群流失的原因與無法提供完整治療，疾病症狀不明顯或與民眾對治療效益的看法改變有關（如更年期）有關。 就病人區隔分析及交叉服務的相關分析都可以發現，以慢性疾病群為中心，針對不同疾病群發展網路治療團隊，應該是未來診所擴張時需要遵循的最重要策略原則；另外健保案件類別的交叉分析，也發現增加預防保健服務可以增加慢性病人的案件，診所需要將成人健康檢查業務當作策略性業務，加強重視並提升品質。 本研究針對描述患者求醫行為過程所發展出對個人主要疾病診斷碼的歸戶處理、RFM相關指標方式、健保卡卡序計算忠誠度及邊際利潤的計算方式對於類似研究應該有其參考價值。至於本研究所提出的診所病患群分群架構，則有待進一步評估其達到CRM顧客最佳化的效果。 / At present, there are much of researches of Customer Relationship Management ( CRM ) and data mining in Taiwan. There is little research in medical service. Our research tried to integrate the three domains knowledge, DBA, domain knowledge of medical service and data mining techniques. This is a case study type research. The CRM Strategy Planning for Outpatient in Kai-Nan Group Practice Clinic by Data Mining on National Health Insurance Dataset. This research included 9300 cases of Kai-Nan Clinic, with nearly 50,000 records of OPD records from Dec, 2000 to Jan, 2003. Our research questions include as followings: 1、How to segment the outpatient, which segment is the most profitable? Which segment is loosing? Why? 2、Which cross service is necessary for what kinds of patients? What kinds of services will be benefit for recruiting chronic patient? What is the reason for the diabetes patient will receive funds examination in this clinic? Are the patients of GYN/OBS will also to be patients of other specialty? Is it necessary to include other specialty in this clinic? 3. Where is the most profitable patient in nearby area? Our study revealed that the most profitable patients is characterized by high medical cost, high loyalty to this clinic, high visit frequencies, high portion of medication fee and high portion of fee for chronic disease. Most of the profitable patients are suffered with chronic diseases. This implies that how to satisfy chronic patient with high satisfaction and earn their trust to be health gate keeper for this patient is very import issue for a clinic. From the results of area analysis for these chronic patients, we concluded the three effects for different areas, such as near-distant effect, public praise and addict effect for original doctors. The segments of chronic patients include golden regular customer、forgetful regular customer、loosing old customer、regular but lower loyalty profitable new customer、irregular general customer、loosed old customer、loosed past customer and loosed old customer. Regarding the segmentation of outpatients of acute illness, we recommended simplify classification according to loyalty that was calculated from the sequence of national health insurance card used in Taiwan. The chronic patients loosed in the clinic was due to lack of comprehensive treatment options, non obvious symptoms or the fears of treatment side effects announced from public media，such as hormone replacement therapy for post menopausal syndrome. We conclude that multidisciplinary team for comprehensive disease management is very important for clinics as our previous success experiences on diabetes patients. Our clinic should expand teams with out bond member according to the needs of our profiles of chronic patients. From association mining, periodic health examinations increase the base of chronic patients. It is strategic important to enhance the staffs and facility for handling periodic health examinations. Our research will also contribute to the following research issues , such as how to describe patients behaviors, how to extract the dominant diagnosis from patients health insurance records, modified RFM dimensions indexes、loyalty based on sequences of health insurance card in Taiwan and the model of calculation of marginal revenue for clinics. As regarding the efficacy of the patients’ segmentation model deserved further study.

傳統上透過行銷研究，如設計問卷透過市場調查、問卷調查等等活動來取得消費者資訊，但在資訊化的時代，利用POS系統所收集到且存在於資料庫中的銷售資料，忠實紀錄所有消費者的消費行為，其中亦隱含著消費者對品牌忠誠度、產品偏好性、價格敏感度等等消費行為。藉由著這些基礎，可進一步探索顧客輪廓(Customer Profile)、顧客忠誠度(Customer Loyalty)及顧客保留率(Customer Retention)，同時預測未來各種行銷活動的可能結果；進而擴大其應用範圍，進行新客源、新產品的開發工作。本研究以資料庫行銷、資料探勘等等理論基礎、統計學技術等等方法，以居家用品業為例，進行資料庫行銷(Database Marketing)的實證工作。實證資料來源為「生活工場」連鎖居家用品店現有之基本資料。並以此資料為基礎，探究顧客購買的消費特性，並以其類別、系列分析其產品關聯性審視其產品組合類別。另外一方面，利用資料庫中的銷售資料及VIP基本資料，以RFM(Recency, Frequency, Monetary)顧客分群方法，分離出不同的顧客族群，並在每一個族群中作交差比對(Cross-Checking)，探究各族群的顧客輪廓及其消費特徵。利用門市基本資料、銷售明細資料等，分析出地域消費特徵。最後，再以金卡、貴賓卡、生活卡三大族群的消費特性及與RFM族群之交差比對，並計算出不同卡別的顧客終身價值(Customer Lifetime Value, CLV)，以期提供「生活工場」與國內連鎖零售體系之產業，未來在對顧客關係經營之參考。 / In tradition, marketing research acquires the information of consumers through questionnaire design, marketing survey, etc. However, in the information age, the POS system is capable to collect the sales information and record them in the data base. It precisely records the consumer behaviors which include the brand loyalty, product preference and price sensitivity. Based on the information, it could more deeply discover the customer profile, customer loyalty and customer retention. It can also forecast the possibility of marketing event for the future and expand to other applications such as new customer creation and new product development. This research, based on the theory of database marketing, data mining methodology and statistic technology, substantiates database marketing in a case study of a home center. Actual data are provided by Working House, a home center chain store. Those data bases establish a foundation of exploring the consumer purchasing characteristic, analyzing series of product classifications and examining the product association and combination. On the other hand, based on the sales information and VIP database, RFM (Recency, Frequency, and Monetary) is utilized to cluster customer segments and research the customer profiling and shopping characteristic by cross checking each shopping group. Data from store locations and sales information are thus employed to explore the geographic characteristic of shoppers. In conclusion, based on the usage data of Gold Card, VIP Card, and Working House Card and cross checks with RFM clustering groups, the Customer Lifetime Value (CLV) of each card will be calculated and accumulated to provide Working House and other retailers the reference for managing customer relationship in the future.

運用資料來幫助企業做出正確且適當的政策是一個存在已久的觀念，在傳統統計上我們通常會將拿到的資料庫直接去作分析，然而對資料採礦(Data Mining)來說，常面臨資料不夠的瓶頸，亦導致資料庫的價值往往不夠。若，我們能利用調查的樣本，推估出目標資料庫中所欠缺的欄位在調查樣本中與其它欄位的關係，便可回推至目標資料庫將原本所欠缺的欄位補齊，將資料庫加大，亦即資料加值(value-added)，那麼，未來要用到這些欄位來分析資料時只要抽樣進行分析即可，如此，也可有效降低企業的成本支出或浪費。 本研究之目的在於整合過去各學者所提出之統計理論與方法，找出誤差指標及模型準則來說明擴充的欄位是有可信度的。由於在目標資料庫擴充欄位時，會產生誤差值，而誤差值的大小往往會影響我們用來判斷此擴充欄位的可行性及可信度，因此本研究並不考慮使用何種抽樣方法，而是假設在簡單隨機抽樣下來進行探討，判別在資料加值前後所造成預測值與實際值之間的差異情形，進一步來做比較。針對欲擴充目標的欄位型態分為連續型和類別型來尋找適當的指標及準備作為我們選擇判斷的指標。類別型欄位利用相似性觀念建立判斷指標，連續型欄位則利用距離觀念、相關性的架構下來討論，如此，可建立合理的誤差指標及模型準則針對欲擴充目標欄位的型態來判斷其擴充的欄位是否具有可信度，並評估其可用價值的高低。 本研究實證結果發現資料庫加值為一可行的方法，從推估資料帶入模式後所得預測值與原始觀測值間計算其相似度皆在九成以上，說明擴充的欄位是有可信度的。 關鍵詞：資料採礦、資料加值、誤差指標、模型準則、相似性 / In recent years, the application of data mining has received good credits and acceptances from a variety of industries such as the finance industry, the insurance industry, and the electronics industry and so on for its success in extracting valuable information translated to opportunities from the database. Database value-added is a new idea not yet fully mature. Its applications on the different databases will have different effect, therefore, the goal of this research is to find the valid and accountable model criteria as a mean to determine if the added columns make any improvement to the database, hence the overall results in terms of predictions. After selecting the model based upon its appropriateness to the data type, we applied the error index and model criteria to evaluate for the performance of the model, if the model has accurately predicted the added-value column. The criterion used in this research is RMSE for the continuous data type and F-value for the discrete data type. Our findings in this research support our attempts that the error index and model criteria used in this research do give us an accountability measure in determining the reliability of adding the columns to the database. Keywords: Data mining, Database value-added, Database, Error index, Model criteria