Chung Cheng University Institutional Repository
Not a member yet
889 research outputs found
Sort by
運用德爾菲法探討巨量資料分析成功之影響因素－以全民健保資料庫為例;Using Modified Delphi Method to Investigate Successful Factors of Big Data Analysis: An Example of the National Health Insurance Database in Taiwan
No full text[[abstract]]巨量資料分析的相關研究與應用，近幾年在學界和業界已愈顯重要。台灣全民健保資料庫(National Health Insurance Research Database, NHIRD)的建置，無疑為醫學研究注入新元素。如何利用巨量資料加強統計支援決策功能及增進學術研究能量，提供政府決策制定與評估之依據已成為近年熱門研究議題。本研究以巨量資料分析專家作為研究對象，針對NHIRD此議題的成功影響因素進行探討，協助各界瞭解進行巨量資料分析研究時應注重之成功影響因素。本研究結合巨量資料分析、資訊系統和商業智慧成功的影響因素，發展出研究模式，以巨量資料特性6V及科技組織環境理論(TOE Framework)三構面等共四大構面，共計14個評估層面及55個問項。本研究採用修正式德爾菲法(modified Delphi)作為本研究的研究方法，遴選國內熟悉巨量資料分析領域的學術界和醫療界、產業專家及政府部門等專家學者參與本研究的修正式德爾菲法進行。本研究共進行二回合問卷，第一回合有30名專家，共計回收23份問卷；第二回合以第一回合回收的23位專家作為問卷發放的對象，此回合最後則回收21份，回覆率則為91%。經過一致性和穩定性檢定後，研究結果顯示出前九大巨量資料分析成功影響因素依序為：定義問題、具價值的資料、資料來源端與組織間的信任、資料解釋的偏見、具備對資料庫熟悉度、資料授權和軟體工具、客戶(個人)隱私、維護資料品質的能力、高度熟練技術的人力資源。而測量的基本問題和結構資料中的有效性與可靠性、安全資料共享、培訓相關職能的人員、擁有必要的工具、及培訓與支持等因素則併列第十名。
The applications of big data analysis have become one of most important topics in academia and industry. However, successful factors of big data analysis effectively support decision making is still unknown. Additionally, the establishment of the National Health Insurance Research Database (NHIRD) undoubtedly injects new elements into the medical research. Therefore, this study aims to investigate the successful factors of big data analysis with an example of the NHIRD.This study develops the research model based on the 6V characteristics of big data and technology-organization-environment framework. Fourteen assessment levels and 55 factors were selected to examine. The modified Delphi method was used in this study. Scholars, industry and government experts in the BDA field were recruited to participate in the study.Twenty-three respondents were received from 30 experts in the first round. Twenty-three experts who responded in the first round were contacted in the second round and finally 21 respondents returned. The response rate was 91%. After the consistency and stability tests, the analytical results indicated that the critical success factors include: clear problem definition, valuable information, trust between the source and the organization, appropriate data interpretation, sufficient database familiarity, data authorization and software tools, customer (personal) privacy, the ability to maintain data quality, highly skilled human resources, measurements of variables, validity and reliability, data security, staff training and resource sufficiency
建構中風及暫時性腦缺血發作病人再入院的預測模型;Developing the prediction model for readmission in patients with stroke and transient ischemic attack
No full text[[abstract]]中風後再入院增加了整體醫療花費支出。了解中風後再入院的原因，有利於照護資源分配及品質改善計劃。 本研究以台灣某大型區域醫院之中風登錄資料中剛住院的資料進行資料探勘，發展預測模型來預測急性中風住院病人的再入院或死亡，以建立其預測模型及探討其預測因子。 共收錄了4197位中風住院病人包含缺血性中風(n = 3165)，腦內出血(n = 445)及暫時性腦缺血發作(n = 587)。14天，30天，90天再入院或死亡的比率，分別為5.92%，9.61%，17.59%。在各時期再入院或死亡的預測模型中整體而言，C4.5或分類與迴歸樹的鑑別力最佳，而邏輯斯迴歸、隨機森林、支援向量機的鑑別力次之，多層感知器或k最近鄰居法的鑑別力最差。CART-風險樹顯示於各時期最高機率再入院或死亡的病人子群如下，14天為中風嚴重度執行命令困難併癌症；30天為中重度中風嚴重度併重度腎功能不佳且高尿酸血症；90天為鼻胃管灌食併輕中度腎功能不佳且肝功能不佳。影響各時期再入院或死亡的共同重要變數，包含過去一年曾住院次數、中風嚴重度、鼻胃管使用、風濕性心臟病、中風嚴重度執行命令。對於這些高風險族群的病人儘早進行出院計劃及照護轉銜介入可以減少再入院或死亡。
Readmission after stroke increases overall medical costs. The knowledge of the risk of readmission and its causes is essential for healthcare resource allocation and quality improvement planning. By using data mining techniques on initial-admission data from the stroke registry of a large regional hospital in Taiwan, this study aimed to develop prediction models and to explore the predictive factors for readmission or mortality in patients hospitalized for stroke. A total of 4197 stroke patients were hospitalized for ischemic stroke (n = 3165), intracerebral hemorrhagic (n = 445) and transient ischemic attack (n = 587). The 14-day, 30-day, 90-day rates of readmission or mortality were 5.92%, 9.61%, and 17.59% respectively. Among prediction models for readmission or mortality within each period, C4.5 and classification and regression tree (CART) methods had the highest discrimination, followed by logistic regression, random forest, and support vector machine techniques, whereas multilayer perceptron and k-nearest neighbor methods performed worst. Based on the CART classifier, patients with increased National Institutes of Health Stroke Scale (NIHSS) 1c (level of consciousness commands) score and cancer, those with moderate to severe NIHSS plus severe renal impairment and hyperuricemia, and those with nasogastric tube feeding plus mild to moderate renal impairment and hepatic function impairment carried the highest probability of readmission or mortality at 14 days, 30 days, and 90 days, respectively. Important predictive factors for readmission or mortality within the three periods included prior admission within 1 year, NIHSS score, nasogastric tube feeding, rheumatic heart disease, and NIHSS 1c score. Early hospital discharge program and transitional care intervention can be targeted for those patients with high risk of readmission or mortality
揮發性記憶體與非揮發性記憶體之混合架構研究與探討;Survey of Hybrid Main Memory
No full text[[abstract]]在現今電腦中，節省電源已經成為了一個很重要的議題，大規模電腦中心與數據中心導致記憶體有日漸變大的趨勢，除了記憶體本身成本增加，記憶體的電源消耗也成為了重大的問題。現階段技術是希望藉由混合揮發性記憶體與非揮發性記憶體，使用它們的優點去彌補對方的缺點。本篇論文研究與探討了數篇關於混合式記憶體的架構與方法，將它們較為成功的混合技術加以討論，並提供它們的實驗結果作為佐證。我們盡量簡要介紹、並全力的探討混合性記憶體的架構及省電還有它們的性能表現。關鍵詞：混合式記憶體、揮發性記憶體與非揮發性記憶體
In today's computer, power saving is become a very important issue. The trend of the memory is getting lager because of the large computer center and data center. In addition to increasing the cost of its memory, the memory power consumption has become a major problem. Modem technology is hope that hybrid volatile memory and non-volatiole memory, use their strengths to make up for shortcomings.In this paper, we survey several paper on hybrid memory architecture and method, and discussed the most successful hybrid technology, then provide their result as evidence. We introduce it short and discuss of hybrid memory architecture 、power saving and their performance.Keyword：Survey、Hybrid Main Memory、Survey of Hybrid Main Memor
輕量型微控器之微架構設計 及其在FPGA上之自適電壓調整實作;Microarchitecture Design of a Lightweight MCU and its AVS Implementation on FPGA
No full text[[abstract]]以重複取樣 （Double Sampling; DS）或抖動偵測 （Transition Detection; TD）為基礎之前瞻臆測 （Speculative Lookahead; SL）技術已被證明在單純的資料路徑設計中優於傳統的Razor原位時序錯誤偵測器 （In-Situ Timing Fault Detector）設計。本論文則進一步探索其在微控制器設計之優勢。首先，我們實作目前最熱門的ARM Cortex M0+微控制器資料路徑，並分別加入Razor、SL和SL/TD三種原位時序錯誤偵測器。由於SL及SL/TD使用雙套資料路徑避免傳統Razor之短路徑 （Short Path）問題，在28奈米製程下可操作在5.75ns的設計，SL比Razor多用了18.8%之晶片面積，而SL/TD只有此設計比Razor少用了1.2%之晶片面積，在其他設計就會多用，但皆可分別省下37.1%和33.7%之功耗。另外本論文亦將含有原位時序錯誤偵測器之ARM Cortex M0+微控制器實作於可調整電壓之FPGA設計平台。以設計軟體估算之時脈的即時JPEG解碼應用可降低21.0%操作電壓，省下44.1%功耗。
Double sampling (DS)- or transition detection (TD)- based speculative lookahead has been proved to have better performance than traditional Razor-like in-situ timing fault detectors in simple datapath designs. In this paper, we further explore SL’s and SL/TD’s superiority in microcontroller datapaths. First, we have designed and implemented ARM Cortex M0+ datapaths, which is the most popular in embedded systems and incorporate Razor, SL and SL/TD in the datapaths for comparison. Due to the duplicated datapaths in SL and SL/TD, 18.8% area overheads and 1.2% area decreases exist in 28nm CMOS compared with that equipped with Razor. However, SL and SL/TD have 37.1% and 33.7% power reduction respectively. Finally, the ARM Cortex M0+ datapaths has been implemented on an FPGA platform that supports dynamic voltage scaling. The supply voltage and power consumption can be reduced by 21.0% and 44.1% respectively for real-time JPEG decoding
基於OpenFlow行動網路之換手機制;Handovers in OpenFlow-based Mobile Networks
No full text[[abstract]]LTE (Long Term Evolution)是一種創新的行動通訊存取網路，它可以提供各種即時和非即時的服務給使用者，以及快速的資料傳輸速率。然而，對於存在於LTE/EPC架構中的網路元件來說，他們大部分都必須要同時處理控制訊號以及資料收送，資料傳輸就需要牽就於處理control signal，如此會導致它們的傳送效能相對低落。為了解決上述問題，許多人認為Software-Defined Networking (SDN)技術是最佳的辦法。SDN可以增加網路資料傳輸效率以及解決網路元件的效能瓶頸。在此篇論文中，我們提出一個基於SDN的LTE/EPC網路架構。我們將SGW和PGW的功能合併為單一顆的實體，如同近幾年有人所提出的LTE/EPC聚合網路一般，我們將合併後的實體稱為S/PGW，如此能達到減少及維護成本的效果。透過SDN的概念，我們將S/PGW分為控制層的S/PGW-C以及傳輸資料的S/PGW-D。此外，我們將S/PGW-C和MME作為應用程式，並執行在controller(此架構中稱為Mobile Controller)上。基地台的部分，在我們系統中的eNodeB也會支援OpenFlow protocol，如此能夠和Mobile Controller進行訊息溝通。Controller方面，我們系統中的Mobile Controller是強化的SDN controller；Mobile Controller除了處理application和控制網路資源外，也會儲存S/PGW-C和MME常會用到的共通資料，像是MPLS的database和UE的位置資訊。此系統我們稱之為OpenFlow-based Mobile Network (OFMN)。另一方面，由於目前OpenFlow協定尚未支援GTP(GPRS Tunneling Protocol)的緣故，我們用MPLS(Multi-Protocol Label Switching)來取代GTP中的資料傳送功能。我們也設計了OFMN中的四種換手流程，分別是OFMN內部換手(包含X2換手或是沒有X2支援的換手)和跨系統的換手(包含從OFMN換至LTE或是從LTE換至OFMN)。這些換手流程能夠幫助UE轉移服務它的基地台而不會造成傳輸中的資料中斷。最後，我們計算了OFMN換手時的控制訊號附載，並將之與其他LTE/EPC架構做比較。我們也計算了各個系統換手時的總延遲(total latency delay)。在這些方面看來，我們可以得知OFMN的表現皆優於其他的比較對象。
LTE (Long Term Evolution) is an innovative mobile access network for various real-time and non-real-time services and high-speed data transmission rate. However, in LTE/EPC (Evolved Packet Core) architecture, almost every network elements need to deal with control signaling and data forwarding simultaneously. It results in lowering data transmission efficiency.Last few years, the new rising technology, Software-Defined Networking (SDN), is considered as a key solution. SDN can improve data transmission efficiency and solve network elements bottleneck. To deal with the increasing loading, we propose a new architecture by applying SDN into LTE/EPC network. In our proposed, SGW and PGW are merged into a single S/PGW, which is the same as LTE/EPC Converged Network. Operators can save equipment expenditure by this separation. We also decouple S/PGW into control plane (S/PGW-C) and data plane (S/PGW-D). S/PGW-C is responsible for handling control signaling while S/PGW-D is in charge of data delivery. Moreover, S/PGW-C and MME are realized as applications running on Mobile Controller. The base station also supports OpenFlow protocol and communicates with Mobile Controller by OpenFlow. In terms of Mobile Controller, it is an enhanced SDN Controller which stores the common information in MME, SGW, and PGW originally, such as MPLS database and UE location. We name our proposed as OpenFlow-based Mobile Networks (OFMN).In addition, we change the way of data routing by replacing GTP (GPRS Tunneling Protocol) tunnel with MPLS (Multi-Protocol Label Switching) because OpenFlow protocol cannot support GTP. We also design four types of handover procedures, including intra-OFMN handover (i.e., X2-based handover and handover without X2 interface) and inter-system handover (i.e., handover from/to LTE network). By means of intra-OFMN handover, UE can handover between base stations in OFMN. OFMN is also able to assisting UE in doing handover from/to LTE architecture through inter-system handover. Our design can maintain the data transmission without interruption after performing these four types of handover.We evaluate the control signaling cost of each handover procedure in OFMN architecture, and compare it with 3GPP and others SDN-enabled architectures. We also use the queuing model to calculate the total latency delay time of each X2-based handover procedures for these architectures. We can see that OFMN architecture’s performance of X2-based handover procedures is better than the others
於低能見度環境之車輛碰撞預測;Vehicle Collision Prediction under Reduced Visibility Conditions
No full text[[abstract]]追尾碰撞經常伴隨著嚴重的交通事故，在智慧型運輸系統中，基於雷達感測的警示系統常受到非直視性或外在環境造成假警報，例如：下雪。近年來，基於車間通訊之車輛追尾碰撞警示系統逐漸發展，以解決上述問題；然而，現有的警示系統仍有很大的空間需要改進，例如：需要適當考慮人的因素；人的因素是影響碰撞警示系統之關鍵，駕駛者的能見度在不同的外在環境下有所不同，而駕駛的反應時間也會隨之變化，進而影響碰撞警示系統的準確度。因此，為了提高碰撞警示系統的準確度，我們提出了基於能見度之碰撞警示系統(Visibility-based Collision Warning System，簡稱 ViCoWS)。這個系統包含四個主要模型，分別為預測範圍估計模型、車速預測模型、車間距預測模型及追尾碰撞警示模型，我們希望透過歷史車速資訊預測未來車速，並且能隨著外在環境動態改變預測範圍，迅速因應即時的交通狀況，以達到提高碰撞警示系統的準確度。實驗結果顯示，我們的車速預測模型平均絕對百分比誤差 (MAPE) 可低於11\%，與前方碰撞機率指標 (Forward Collision Pribability Index，簡稱FCPI) 相比，在交通順暢且低能見度的情況下，ViCoWS可提早4.5秒和2.1秒發出碰撞警告，而FCPI可提早0.6秒發出碰撞警告；而在交通壅塞且低能見度的情況下，ViCoWS可提早1.9秒發出碰撞警告，而FCPI可提早1.2秒發出碰撞警告。
Rear-end collisions always cause serious traffic accidents. In intelligent transportation systems (ITS), radar collision warning systems are highly accurate in determining the range and in detecting the rear-end of a vehicle; however, in poor weather such as fog, rain or snow, the accuracy is significantly affected. In recent years, the development of Vehicle to Vehicle (V2V) or Vehicle to Infrastructure (V2I) communication systems for traffic safety issue has been popular in the field of ITS to deal with the problem. However, there are still much left for improvement. For example, weather conditions always impact on human factors, which is one of the important factor on collision warning algorithm. Then, the accuracy of CWS is affected.For increasing the accuracy of CWS, we propose a Visibility-based Collision Warning System (ViCoWS). This method contains four main models, including prediction horizon estimation model, velocity prediction model, headway distance prediction model, and rear-end collision warning model. Historical velocity data is used to predict future velocity volumes. A prediction horizon model is proposed to change prediction horizon in relation to weather conditions. It can rapidly respond to real-time weather conditions to increase correct warning.Experiment results show that the mean absolute percentage error of velocity prediction model is less than 11%. In traffic unhindered under low visibility conditions, ViCoWS can warn the driver by about 4.5 second prior to the collision and 2.1 second prior to the collision, where the Forward Collision Probability Index (FCPI) method gives 0.6 second for the driver to react before collision. In traffic congestion under low visibility conditions, ViCoWS can warn the driver by about 1.9 second prior to the collision, where the FCPI method gives 1.2 second for the driver to react before collision
資訊工程學生之創造力與學習環境因素分析;Analysis of Creativity and Leaning Environment Factors with Computer Science and Information Engineering Students
No full text[[abstract]]本研究主要探討學習環境對資訊工程領域學生想像力的影響，並且建立環境影響因素對想像力的假設模型，從中得知各個環境因素對於資工系學生的想像力是否受到直接的影響，同時考慮其中是否有中介影響的存在。在研究當中，我們對於資訊工程領域學生發放問卷調查，共收回161名有效樣本，題項主要由促發想像的四個環境因素組成:社會氣氛、學習資源、組織方法、文化特質等。每位學生都必須具備專題實作或論文撰寫的經驗，再根據自身經驗填答問卷。研究分析的方法分別利用SPSS 18.0及AMOS 18.0進行樣本的敘述性、探索性、驗證性等因素分析，再利用效度與路徑分析來探索假設模型是否成立，最後討論模型路徑中的中介效果與交叉比較。在研究過程中，假設的模型遇到了效度情況不佳的狀況，使得模型中某些假設路徑因而被刪除，而也藉此發現了其他的假設路徑，而建構了新的假設模型，新的假設模型雖在驗證性分析為勉強符合標準，但總體而言，新的假設模型在效度及路徑分析上都是顯著的。研究結果顯示，文化特質對於組織方法和學習資源具有正面的影響，而文化特質也能同時透過學習資源間接影響組織方法。在交叉比較方面，大學生較偏向於在學習資源和組織方法，而研究生較注重學習資源和文化特質。
This Thesis mainly discusses the influence factors of learning environment on the creativity of students in the field of Computer Science and Information Engineering (CSIE). We have established a hypothesis model of the environmental factors influencing creativity. We investigate whether the environmental factors have a direct influence on the imagination of the CSIE students, and find out whether there is any intermediate influence. In this work, an online questionnaire survey was conducted on CSIE students in National Chung Cheng University. A total of 161 valid samples were collected. The main questionnaire in the learning environment were deconstructed into four factors: Human Aggregate, Organizational Measure, Social Climate, and Physical Component. Each student must have a graduate project or thesis writing experience, and then fill the questionnaire according to their own experience. The methods of analysis using SPSS 18.0 and AMOS 18.0 to do the descriptive, exploratory and confirmatory factors analysis. The validity and path analysis were used to explore whether the model was established. Finally, the intermediate effect and cross comparison were discussed. In the course of the study, our proposed model encounters a poor situation in the step of validity analysis, so that some of the hypothetical paths in the model are thus deleted, and the other hypothetical paths are found. Finally, a new hypothesis model is constructed. Since the case of confirmatory analysis of new hypothesis model reluctantly meets the criteria, but overall, the new hypothesis model is significant in validity and path analysis. The main result shows that Human Aggregate has a positive effect on Organizational Measure and Physical Component, and Human Aggregate also indirectly affects Organizational Measure through Physical Component. In the cross comparison, the undergraduate students tend to be affected by the Physical Component and the Organizational Measure, while the graduate students are more affected by the Physical Component and the Human Aggregate
植基於BI-RADS之乳房超音波影像 多類別電腦輔助診斷系統;A Multi-class Computer-Aided-Diagnosis System of Breast Ultrasound Image Based on BI-RADS Categories
No full text[[abstract]]乳房超音波不僅是乳房癌症檢測中對人體影響較小的非侵入性腫瘤篩檢儀器，也是乳房腫瘤初步篩選的重要方法之一；儘管如此，不同機種之乳房超音波影像之間常會存在著色彩差異、解析度等成像問題，造成臨床醫師難以詳細判讀腫瘤良惡性或等級。 本篇論文採用三種超音波機種( PHILIPS、SIMENS、TOSHIBA ) 乳房影像作資料樣本集，然後針對樣本影像實施影像標準化、對比強化以及腫瘤分割等步驟；並延伸傳統乳房腫瘤嚴重性分類，進一步將良惡性嚴重程度根據美國放射學會之乳房影像報告與資料系統(breast imaging-reporting and data system, BI-RADS)，針對臨床需求大之BI-RADS 2 ~ 5等四個嚴重程度等級，擷取出與病灶嚴重性分類之相關特徵，配合三個機器學習分類器(支持向量機、隨機森林、卷積神經網路)與特徵選取動作，進行腫瘤嚴重性之多類別綜合性評估，藉此判斷出所病例中乳房腫瘤真正嚴重程度。本系統在BI-RADS樣本判斷的結果，SVM的正確率有80.00%，RF的正確率有77.78%，CNN的正確率有85.42%；此外本研究還進一步確認各種超音波成像儀器在BI-RADS分類的適應性，發現使用CNN進行BI-RADS分類的F-score 會達到75%以上，也就是CNN有不錯的分類效能與適應性。 我們期望所研製出之植基於CNN之電腦輔助診斷系統能輔助臨床醫師判讀，讓病患盡快接受合適的治療或者避免不必要的手術。
Ultrasound image is not only a non-invasive instrument of less impact on human body for breast carcinoma detection, but also a basic tool to detect breast tumor. However, there exist some problem such as color difference and different resolution in image acquisition among different types of ultrasound imaging modalities so that clinicians always can’t identify accurately the BI-RADS categories or disease severities. In the study, three types of breast ultrasound imaging modalities including PHILIPS, SIMENS, and TOSHIBA were adopted to fetch breast ultrasound images for our experimental samples. Then, processing stages such as intensity normalization, contrast enhancement, and image segmentation were performed sequentially to detect true breast tumor. Our proposed system identifies the breast tumor severities according to the Breast Imaging-Reporting and Data System (BI-RADS) rather than traditional assessment on the severity, i.e. merely using benign or malignant. After segmentation, we focused on the BI-RADS 2-5 due to the clinical practice. And, several features related to lesion severities based on the selected BI-RADS categories were fed into three machine learning classifiers including Support Vector Machine, Random Forest, and Convolution Neural Network combined with feature selection to develop a multi-class assessment of breast tumor severity based on BI-RADS. Experimental results tested on BI-RADS samples reveal that the identification accuracies with SVM, RF, and CNN are 80.00%, 77.78%, and 85.42, respectively. In order to validate the performance and adaptability of classification using different ultrasound imaging instruments, the evaluations of F-score based on CNN obtain measures higher than 75% (i.e., prominent adaptability) when samples were tested on various BI-RADS categories. We hope that the Computer-aided diagnosis (CAD) system developed based on CNN can provide diagnostic references for surgeon in interpreting BI-RADS categories so that patients can obtain appropriate treatments or avoid unnecessary surgeries