Application of Conversion Degree Function and SVM Regression in Precipitation Forecasting
Lili Chen, Hongjun Guo · International Journal of Applied Mathematics & Statistics/International journal of applied mathematics and statistics · 2013
For the purpose of analyzing and developing precipitation forecasting, the influence of historical factors on the prediction is fully considered. For the model of precipitation sequence with seasonal law, time series of precipitation are divided into a number of data windows with equal widths, and the similarity search is performed on them to find a data window whose running track is the most similar to that of the current time series’ according to the qualitative mapping and conversion degree function in attribute theory. When multiple similar data windows are presented in searching results, the secondary judgment will be made by the weighted sum model with variable weight based on attribute coordinate system to find the most similar data window which can be used for precipitation forecasting. If there are no historical factors for reference, the prediction will be made by using support vector machine regression to put precipitation time series fitting out.