基于机器学习方法的贵阳雾预报模型研究

Research on Fog Forecasting Model Based on Machine Learning Method

选取2015—2018年贵阳国家气象站逐时观测数据,归纳整理了影响雾、浓雾的11个气象因子,在分析影响因子与能见度关系的基础上,以多次随机选取80%的样本数据为训练集,剩余20%的样本数据为预报模型的检验集,利用机器学习中C5.0、CART和神经网络算法分别构建贵阳站雾和浓雾的预报模型,检验评估了各预报模型的应用效果。结果表明:基于机器学习的雾、浓雾预报具有较好的业务应用前景,3种算法的检验准确率均在90%以上,其中CART算法对贵阳站雾的预报效果最好,C5.0算法和神经网络的多层感知器算法对贵阳站浓雾的预报效果最佳。

Based on the hourly observation data of Guiyang National Meteorological Station from 2015 to 2018,11 meteorological factors affecting fog and dense fog were summarized.Based on the analysis of the relationship between influencing factors and visibility,80% of the sample data were randomly selected as the training set,and the remaining 20% of the sample data were used as the test set of the prediction model.The prediction models of fog and dense fog in Guiyang station were constructed by using C5.0,CART and neural network algorithms in machine learning,and the application effects of each prediction model were tested and evaluated.The results show that the fog and dense fog prediction based on machine learning has a good business application prospect,and the test accuracy of the three algorithms is above 90%.Among them,the CART algorithm has the best prediction effect on the fog at Guiyang station,and the C5.0 algorithm and the multi-layer perceptron algorithm in the neural network have the best prediction effect on the dense fog at Guiyang station.