Construction of a Clinical Decision Support System Using Ensemble Classification
| Autor: | Wu, Sheng-Han, 吳昇翰 |
|---|---|
| Rok vydání: | 2013 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 101 Clinical Decision Support System (CDSS) assists clinical staff in the diagnoses of diseases, provision of information for care support, and improvement of efficiency to enhance the quality of care. Accurate data analysis is the crucial element during the diagnosis process. Nowadays novel techniques have been developed to provide a wide range of data analysis and applied in various fields to obtain valuable information. However, single classifier does not perform consistently for all data sets. Integration of recommendations and proposed measures for improving classification effectiveness were suggested to avoid the inconsistency. In this research, Genetic Algorithm (GA) was combined with Support Vector Machine (SVM) to provide a foundation for CDSS design. The proposed EnsCV method was compared with BAIS based on the data in the UCI machine learning database containing 11 datasets. It was shown that EnsCV is more effective, especially for Sonar and Glass, in the classification of categorical datasets. For instance, the accuracy using the proposed EnsCV method is 6.9% and 13.3% higher than the BAIS in classifying the datasets Sonar (EnsCV: 93.3%; BAIS: 86.4%) and Glass (EnsCV: 86.9%; BAIS: 73.6%), respectively. It is concluded that the ensemble classifiers present greater classification performance than the traditional method. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|