Application of SVM networks in hybrid models for forecasting and estimating maximum and minimum daily humidities

Abstract

Daily environmental humidity level forecasting is one of the problems that is concerned not only in Vietnam but also in other countries in the world. The prediction model is highly dependent on geographic and regional conditions. Therefore, in different regions, it is necessary to find the appropriate data sets and models for the forecasting solution. In this paper, we propose to use a hybrid model combining of an SVM (Support Vector Machine) and a linear block for forecasting and estimating maximum and minimum daily humidity values in Hai Duong City, Vietnam. The input data are the historical values of the maximum, minimum of temperatures, humidity, wind speed and mean value of precipitation, the number of sunshine hours. The quality of the proposed solution was tested on the official observation data (2191 days, 01/01/2010 to 31/12/2015) collected by the Central Meteorological at The North Central region of Vietnam for 6 provinces (Hai Duong, Bac Ninh, Thai Binh, Hai Phong, Quang Ninh and Hung Yen). The empirical results show an average error of 3, 35% with the predicted model and 3, 59% with the estimated model.