Time series analysis and prediction of climate variables of Southern Java waters using support vector regression

Southern Java Waters contribute significantly to aquatic economy of Indonesia by providing abundant fisheries resources.The understanding of sea state of the waters become a focal point.The aims of this paper are to analyze and to predict time series dataset comprising climate variables such as wind speed, surface temperature (SST), precipitation, and surface pressure of Southern Java waters.The analysis has been done by decomposing the time series dataset to its patterns, trend and seasonality, and calculating the correlation matrix of the dataset.The prediction method employs support vector regression (SVR) algorithm.The performance of the resulted models is computed using mean squared error (MSE).The result shows that wind speed of Southern Java Waters is positively corelated with surface pressure and negatively corelated with SST and total precipitation.The lowest MSE occurs in SST model.Meanwhile, the largest MSE is the total precipitation model.The developed models could be used as prediction tools of climate variables for following years in the Southern Java waters.