Optimizing Margin in Weighted Support Vector Machines for Flood Problem

Nowadays, the problem of floods has occurred frequently and it is a natural disaster that has been faced by people of the world. Thus to obtain the perfect information to be analyzed and classified accurately is quite impossible. Noise data decreases the quality of the data and can give poor result on the performance of learning algorithms especially on the classification accuracy. All of these limitations had cause the use of classification model for the inaccurate extraction of distribution pattern of floods in the broad areas. In this study, we investigate the drawbacks of classical weighted support vector machines (WSVM) and improve the WSVM algorithm using multiple hyper-plane and Instance-weighted (MHI) for flood classification. We compare our proposed method with existing WSVM to gain better decision margin for higher classification accuracy. The proposed method is evaluated using three KEEL datasets and real flood datasets. The results indicate that the proposed approach can be applied to existing approaches for computation with significantly in adapting and showed the improvement of the classification accuracy.