The Construction of Text Mining and Data Mining Technologies for Forecasting Endometrial Cancer
| Autor: | CHUNG,LING-LING, 鍾玲玲 |
|---|---|
| Rok vydání: | 2016 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 105 The incidence rate of endometrial cancer is the fastest growing cancer and most common gynecological cancer in the last decade. Diagnostic methods and technological improvement establish an organized and systematic approach, it could detect the early cancer and make greater progress in cancer prevention and control. Patient ' s diagnosis and health data storage transfer the traditional paper-based medical records into electronic medical records and serves as the main source of medical information in clinical applications, medical education and investigation. Objectives: Data mining technology has been widely applied in various medical research and the key points from the data could be applied to medical decision making. Therefore, this study aims at (a) the use of Text mining technology to explore the impact of endometrial cancer-related factors. (b) To establish the forecasting endometrial cancer risk model and risk index by using Data mining technology. Methods: In this study, 890 cases of endometrial biopsy were collected from a Regional Teaching Hospital in Chiayi City from 2006 to 2015. Among them,148 cases with ICD-9 code【182】of endometrial carcinoma were the case study. The forecasting model of endometrial cancer was constructed by Decision tree, Support vector machines and Logistic regression. The best performance classification model was evaluated by the performance index. Results：The average accuracy prediction rates of endometrial cancer are as below: Support vector machines model is 96.9%, Logistic regression mode is 95.80% and Decision tree model is 91.80%, meanwhile, generalize the risk tree of endometrial cancer. Conclusion: In this study, we used the records of medical institutions, including: patient’s complaints, physical examination findings, ultrasonography, pathological reports, etc,. By editing, organization and analysis process of Text mining, furthermore established the forecasting model for clinic medicine by the exploit of Data mining .It could make up for the inadequacies of the general statistical analysis and reveal the association between medical records and endometrial cancer, providing clinicians assistance in the assessment of patients as a reference. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|