指導教授：歐陽彥正臺灣大學：資訊工程學研究所陳品良Chen, Pin-LiangPin-LiangChen2014-11-262018-07-052014-11-262018-07-052014http://ntur.lib.ntu.edu.tw//handle/246246/261431近年來，大型臨床醫學資料庫的研究越來越受到關注。傳統的統計與時間序列分析方法經常被使用，但都有其侷限性。因此，發展一個更好的時間序列分析法是必須的。本論文中，我們發展了傅立葉－高斯分解方法，利用傅立葉轉換、高斯函數與一個最佳化演算法，可將一個訊號分解出數個具有不同代表意義的趨勢。傅立葉－高斯分解法可萃取出一個訊號中具有相同頻率的不同趨勢，這是其他方法所做不到的。此外，我們將傅立葉－高斯分解法應用在分析台灣全民健康保險研究資料庫的疾病及到院前心肺功能停止資料庫，並且得到了一些有趣的發現。我們在過敏性鼻炎、氣喘、急性心肌梗塞的看病人次中找到了特殊的趨勢。過敏性鼻炎的看病人次在每年的三月和十一月達到最高峰；氣喘的看病人次在四月和十一月達到最高峰；急性心肌梗塞的看病人次則是在在立春、立夏、立冬達到最高峰。此外，我們發現循環系統疾病與消化系統疾病的看病人次有相同的趨勢。它們都在春分、芒種、冬至大幅下降。在到院前心肺功能停止資料庫中，我們發現非創傷性到院前心肺功能停止的病患人數在冬天大幅增加而夏天小幅度增加。非創傷性到院前心肺功能停止病患的存活率則在春天與秋天上升，與病患人數呈現相反的趨勢。In recent years, more and more studies have focused on the large medical databases. Traditional statistical approaches and time series analysis methods are frequently used, but they have some limitations. Therefore, to develop an advanced time series analysis method is required. In this thesis, we develop the Fourier-Gaussian decomposition method and show that it can decompose a signal into a finite and small number of components. Fourier-Gaussian decomposition can extract different components with the same frequency from a signal, which is not available in other methods. Furthermore, we apply Fourier-Gaussian decomposition to analyze several diseases in Taiwan’s National Health Insurance Research Database (NHIRD) and the out of hospital cardiac arrest (OHCA) database. Finally, we get some interesting findings. We find special patterns in allergic rhinitis visits, asthma visits, and AMI visits. Allergic rhinitis visits contained one-year period and peaked in March and November; asthma visits peaked in April and November; AMI visits peaked in Spring Begins, Summer Begins and Winter Begins. Besides, we find that circulatory system diseases visits and digestive system diseases visits have the same pattern. The number of patients decreased rapidly at Vernal Equinox, Grain in Ear and Winter Solstice. In OHCA database, we find that the number of non-traumatic OHCA patients increased rapidly in winter and slightly in summer. The survival rate of non-traumatic OHCA patients increased in spring and autumn, which is reverse to the number of non-traumatic OHCA patients.致謝 i 摘要 ii Abstract iv Contents vi List of Tables viii List of Figures ix Chapter 1 Introduction 1 1.1 Motivation 1 1.1.1 Development of time series analysis 2 1.1.2 Electronic health record systems 3 1.1.3 Analysis on different time scales 4 1.2 Contribution 5 1.3 Organization 5 Chapter 2 Background 8 2.1 Time series analysis methods 8 2.1.1 Fourier transform 8 2.1.2 Wavelet transform 9 2.1.3 STL 10 2.1.4 Hilbert-Huang transform 11 2.1.5 Other methods 13 2.2 Rank-based adaptive mutation evolutionary algorithm 14 Chapter 3 The Medical Databases 19 3.1 National Health Insurance Research Database 19 3.2 Study samples 21 3.3 Out of Hospital Cardiac Arrest database 23 3.4 Study samples of OHCA database 23 Chapter 4 Fourier-Gaussian Decomposition Methods 26 4.1 The first phase of the FGD algorithm 26 4.2 The second phase of the FGD algorithm 29 4.3 Comparison with other decomposition methods 36 4.4 Strengths and Limitations 36 Chapter 5 Analysis of Medical Databases 40 5.1 Mental disorders 41 5.2 Diseases of the sense organs 41 5.3 Diseases of the respiratory system 42 5.4 Diseases of the circulatory system 43 5.5 Diseases of the digestive system 45 5.6 Diseases of the genitourinary system 47 5.7 Diseases of the skin and subcutaneous tissue 48 5.8 Out of hospital cardiac arrest 49 Chapter 6 Discussion and Conclusion 69 6.1 Discussion 69 6.2 Conclusion 70 6.3 Future work 71 References 721287800 bytesapplication/pdf論文公開時間：2019/08/16論文使用權限：同意有償授權(權利金給回饋學校)傅立葉轉換高斯函數傅立葉－高斯分解法全民健康保險研究資料庫到院前心肺功能停止資料庫時間序列分析thesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/261431/1/ntu-103-D97922006-1.pdf