skip to main content
語言:

B型肝炎e抗原轉態與病毒序列突變率預測之研究

劉威良; Liu, Wei-Liang 孫光天; Koun-Tem Sun; 資訊教育研究所碩士班 2006

線上取得

  • 題名:
    B型肝炎e抗原轉態與病毒序列突變率預測之研究
  • 著者: 劉威良; Liu, Wei-Liang
  • 孫光天; Koun-Tem Sun; 資訊教育研究所碩士班
  • 主題: e抗原; 突變率; 類神經網路; B型肝炎病毒; 分類; 預測; HBeAg; prediction; mutation rate; hepatitis B virus; classification; artificial neural network
  • 描述: B型肝炎病毒感染一直是台灣地區重要的公共衛生課題之一，因此本研究係以成功大學醫學院所提供之B型肝炎患者臨床資料為實驗樣本，並選擇對時間序列較常應用之類神經網路模型稍作改良，期望能以患者過去各月的樣本特徵，透過類神經網路推論，預測B型肝炎患者e抗原轉態及下個月B型肝炎病毒序列突變率，甚至能預測至更久的時間。經過本研究預測結果，在e抗原轉態預測，其敏感度、特異度及平均預測能力分別為74.44%、82.81%、81.73%。而以下個月病毒序列突變率為預測目標，其預測結果均方根誤差最小為0.00175，最大也僅有0.00277；而平均比例誤差最小為0.08449最大為0.32982。最後，以第六個月病毒序列突變率為預測目標，其均方根誤差方面，四位病患除了一位病患0.00381稍高外其餘皆在0.002以內；在平均比例誤差方面，最小為0.245442最大為0.781983。經由上述研究結果顯示，確實以時間序列為概念的類神經網路模型，在預測Ｂ型肝炎病患e抗原轉態及病毒序列突變率皆得到不錯的效果。
    The infection of Hepatitis B Virus (HBV) has always been one of the most significant studies of public health in Taiwan. Thus, this research has taken the clinical information of HBV patients provided by College of Medicine National Cheng Kung University as experimental specimen. Also, a slightly modified model of Artificial Neural Network, which regularly is applied to time series, has been chosen for predicting HBeAg seroconversion of HBV patients, mutation rate of HBV DNA sequence of the next month, and even for making it possible to predict for a longer period of time through referencing Artificial Neural Network based on the patients’ specimen of the past months. The predicted result indicates that the sensitivity, specificity and average prediction ability of HBeAg seroconversion are 74.44%, 82.81% and 81.73%. Besides, if mutation rate of HBV DNA sequence of the next month is taken as a predicted goal, the predicted result shows that the minimum Root Mean Squared Error is 0.00175 and the maximum is only 0.00277; while the minimum Mean Relative Error is 0.08449 and the maximum is 0.32982. Finally, if mutation rate of HBV DNA sequence of the sixth month is taken as a predicted goal, only one out of the four patients is at 0.00381 in terms of Root Mean Squared Error, which is slightly higher, and the patients are below 0.002. The minimum Mean Relative Error is 0.245442 and the maximum is 0.781983. The above result shows that the model of Artificial Neural Network based on the concept of time series has verified its positive effect on predicting HBeAg seroconversion of HBV patients and mutation rate of HBV DNA sequence.
    碩士
  • 建立日期: 2006
  • 格式: 121 bytes
    text/html
  • 語言: 中文
  • 識別號: http://nutnr.lib.nutn.edu.tw/handle/987654321/4335
  • 資源來源: NUTN IR
連結
返回到檢索清單
上一頁前一個記錄 檢索清單  6 下ㄧ個記錄下ㄧ頁
    臺南大學版權所有 © 2016 NUTN All Rights Reserved

正在檢索遠程資料庫，請稍等