Author(s): C. Sivapragasam

Linked Author(s):

Keywords: Support Vector Machine (SVMS); Structural Risk Minimization (SRM); Neural Networks (NN); Singular Spectrum Analysis (SSA)

Abstract: To predict the anomalies of the rainfall time series is a complex task. Owing to the stochastic and unpredictable nature (or chaotic and short term predictability) of the process, rainfall time series, which contains significant noise, has been poorly predicted. This paper investigates the application of a relatively new machine learning technique, Support Vector Machines (SVMs) together with a pre-processing algorithm based on Singular Spectrum Analysis (SSA) in the Singapore rainfall prediction. Whereas SSA decomposes original time series into a set of high frequency and low frequency components, SVM helps in efficiently dealing with the computational and generalization performance in a high-dimensional input space. Pre-processing of the raw data with SSA significantly improved the prediction results. The 1-lead day prediction gave correlation coefficient as high as 0.82 for training and 0.71 for verification.

DOI:

Year: 2001