Support vector classification and back-propagation neural network based models for ground water level forecasting

• Main Authors: Cai-Rong Liou, 劉寀蓉
• Other Authors: Ching-Tien Chen
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/46397074270809915472

碩士 === 國立嘉義大學 === 土木與水資源工程學系研究所 === 100 === Recently, global warming and climate change issues are an important threat to this planet, and the great changes in hydrologic environment are an impact of the climate change caused by continuous warming. In Taiwan, owing to nonuniform temporal and spatial distributions of precipitation and the presence of high mountains and steep channels all over the island, the ground water and surface water resources are not managed efficiently. Moreover, the ratio of the total use of surface water resource to groundwater resource is approximately 7∶3, which shows the importance of the groundwater resource. Groundwater can be the most valuable resource if good forecasts and proper disaster mitigation are made. In this study, based on the combination of the support vector classification (SVC) and the back-propagation neural network (BPN), a groundwater level forecasting model is proposed, named SVC-BPN model. The proposed model is applied to actual groundwater data from 6 groundwater stations of the alluvial fan of the Zhuoshuixi River in southern Taiwan. The results show that the BPN model performs worse. Then, the SVC is applied to binary classification of the groundwater level data. It is found that the average correct classification rate is only 70%, and the BPN model can forecast less precisely than the BPN model without SVC classification. Finally, we assume that if the correct classification rate is 100%, then the overall performance of BPN model is effectively improved. Hence, more studies and discussions will be carried out on raising the accuracy of SVC classification for improving the generalization ability of the proposed model in the future.