| Freezing is the weather phenomena that the temperature is below 0℃in winter, and the cooling and freezing precipitation or solid precipitation and icing phenomena. It is the formation in particular weather conditions, and highly localized. The impact of a wide range, long duration and intensity on transportation, energy supply, power transmission, communication facilities, industrial and agricultural production have a serious influence on people's life. Therefore, the thesis of predicting freezing is particularly important, and the forecasting model selection and research is of great significance.Based on the data of the influence of the main factors - temperature, relative humidity, rainfall, air pressure and wind speed, this thesis used support vector machines for classification power of the question, identified whether there is freezing, to achieve the exact purpose of forecasting freezing, and provide a new idea for it.First, based on 1969-2008 annual data of temperature, relative humidity, rainfall, air pressure and wind speed, this thesis have a statistical analysis on five variables on interannual variation of congeal and spatial distribution; Second, select the 1969-2000 data to Frozen in effect as the test samples, and 2001-2008 data as the test samples, respectively forecast the freezing through the support vector machine ploy kernel function and RBF kernel function, while establishing a freezing prediction model both BP neural network and combination forecast; Subsequent, analyze the data of errors of the support vector machine, BP neural network and combined forecast on forecasting the freezing; Finally, this thesis compared the results the model predicted.The results show that the accuracy of the prediction of SVM RBF kernel function is the highest, compared with the predicting mode of BP neural network and Combination forecasts, and short-term forecasting model for its thesis also more accurate and it opens up a new way for weather forecasting. |
PDF Full Text Request