| Summary: | 碩士 === 國立中正大學 === 資訊管理學系 === 92 === Abstract
Tuberculosis against announced as a critical problem in 1993, World Health Organization (WHO) have established the aim to manage tuberculosis (TB) cure rate. Current global tuberculosis efforts are threatened in Taiwan now, is worried about gradually decrease cure rate and high incidence rate. This study hopes to predict the tuberculosis’s patient arise the ratio of cure rate, and prevent multidrug-resistant TB (MDR-TB) occurs.
We search the med-line from 1963 to 2003; all papers related about tuberculosis and complete treatment, separate the affected factors as independent variables. This research uses 1029 medical records in a regional teaching hospital in Chia-Yi city database as sample, focus on the building the predict model by experiment of data mining technique and linear regression applying on TB patient’s medical database. It’s purposed to use scientific method to judge linear regression and neural networks can provide assured evidence to establish clinical guideline and then promote the quality of medical care.
Under the same condition, the result of neural networks model is better than regression model, and MSE value is also low. Compliance is the major factor in both models.
To sum, it is important to improve the prediction model of TB cure rate by applying data mining technique and implement TB health policy relying on building knowledge management system by CDC.
|
|---|
