Support Vector Machine (SVM) for Rainfall Forecasting at Johor River

Rainfall prediction plays an important part in forecasting early warnings of heavy rainfall and flash floods. In this study, rainfall data from Ladang Getah Malaya, Kota Tinggi at Johor state, Malaysia is taken for the rainfall prediction model over a period of 60 years. The method used to build the prediction model is known as the support vector machine (SVM) method. The results indicate the SVM utilizing the radial basis function (RBF) kernel performed the best among four kernels (RBF, Sigmoid, Linear, and Polynomial). Even though the results were less satisfactory than expected, adjustments could possibly be made to this model in order to improve its performance. Some of the reasons why the degradation of the performance occurred are extremely large values inside the actual data affected the performance of the model and data might not be as accurate as possible due to equipment errors during measurement

Keywords:

Support Vector Machine , Rainfall Prediction , Johor State