skip to main content

MODEL PREDIKSI CURAH HUJAN DENGAN PENDEKATAN REGRESI PROSES GAUSSIAN (Studi Kasus di Kabupaten Grobogan)


Citation Format:
Abstract

Forecasting method of rainfall has developed rapidly, ranging from the deterministic approach to the stochastic one. Deterministic approach is done through an analysis based on physical laws expressed in mathematical form, which identify the relationships between rainfall and temperature, air pressure, humidity and the intensity of solar radiation. Similarly, there are some stochastic models for the prediction of rainfall that have been commonly used, for instances, the model Autoregressive Integrated Moving Average (ARIMA), Fourier analysis and Kalman filter analysis. Some researchers about climate and weather have also developed a predictive model of rainfall based on nonparametric models, especially models based on artificial neural networks. Above models are based on classical statistical approach where the estimation and inference of model parameters only pay attention to the information obtained from the sample and ignore the initial information (prior) of parameter model. In this research, prediction model with Gaussian process regression approach is used for predicting the monthly rainfall. Gaussian process regression uses a stochastic approach by assuming that the amount of rainfall is random. Based on the value of Root Mean Square Error Prediction (RMSEP), the best covariance function that can be used for prediction is Quadratic Exponential ARD (Automatic Relevance Determination) with RMSEP value 123,63. The highest prediction of the monthly rainfall is in January 2014  reached into 336,5 mm and  the lowest in August 2014 with 36,94 mm.

 

Key Words: Gaussian Procces Regression, Covariance Function, Rainfall Prediction

Fulltext View|Download

Article Metrics:

Last update:

No citation recorded.

Last update: 2024-10-31 22:11:23

No citation recorded.