A Forecasting Method with Multivariable Analysis for Prevention of Dengue Outbreaks Based on Simple Recurrent Neural Network Techniques
| Autor: | Chung, Xiang-Hong, 鍾相宏 |
|---|---|
| Rok vydání: | 2017 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 105 Dengue fever was a disease that feels both familiar and fearful for the countries located in the tropical and subtropical region, especially the breeding grounds and territorial expansion of mosquito day by day caused by the climate warming and the development of international trade and traffic. Consider the case of Taiwan, there were serious outbreaks of dengue fever in the south of Taiwan in 2014 and 2015, the number of dengue cases was unprecedented, resulting in many death cases. Although the epidemic of dengue fever has been eased in 2016 and doesn't have serious outbreaks until now, we still need to keep the warning attitude for the future days. Therefore, we discussed a multi-category classifier based on neural network techniques in this work and simulated the prediction of dengue outbreaks through classifying the new data by the classification models. We collected and unified the historical data related the dengue outbreaks during the period from 2007 through 2016, and the data, including infected case, environmental factors, climate factors and event factors in this work. Through the official epidemic level definition combine with the classify method of the Simple Recurrent Neural Network (SimpleRNN) for learning to forecast the class of dengue outbreaks monthly. Although the predicted result couldn’t match in real outbreaks level perfectly, it was still being able to warn the epidemic early one month and remind the related unit started to prevent a kind of level of epidemic outbreaks. In the final, we obtain the experience by the resulting discussion from experiment to improve the prediction method of related research in the future. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|