摘要
道路结冰是气象上规定的主要灾害性天气之一,准确认识道路结冰发生规律并科学有效地预报是防灾减灾的关键问题。以影响道路结冰的日积雪深度、日最低气温、日降水量等气象要素为主要因子,将是否结冰抽象成二元分类问题,借助数据挖掘中C4.5决策树算法进行分类分析,得到便于预报员使用的预报道路结冰规则集。在实验部分,选取2004—2013年连云港和嘉定为例,用前8 a的连云港(嘉定)地区道路结冰数据样本进行分类规则的学习,从而得到适合连云港(嘉定)地区的4(3)条预报道路结冰规则集,然后用剩下2 a的数据进行检验,准确率高达96.82%(94.87%)。实验表明采用C4.5决策树算法构建的道路结冰预报模型不仅预报准确率高、易解释、方便预报员直接使用,而且预报规则与道路结冰现象的天气过程及理论一致。
Road icing is one of the major disastrous weathers in meteorology.Accurately understand- ing the occurring law of road icing and scientifically forecasting it are the keys to effective disaster preven- tion and mitigation. Firstly, daily snow depth, daily minimum temperature, daily rainfall amount are taken as the main meteorological factors affecting road icing, and then the problem whether the freezing happened at last is abstracted into the binary classification one. Secondly, C4.5 decision tree algorithm in data mining is adopted to find the classified roadicing forecasting rules that are easily used by forecasters. Finally, road icing rule set is got by carrying out experiment with roadicing sample of data in Lianyungang and Jiading districts from 2004 to 2013. The first eight years＇ data of Lianyungang regional road icing data samples are used to gain classification rule so as to find four road icing prediction rules fitting Lianyunga- ng district, then the remaining two years＇ data are used to validate with 96.82% accuracy rate. The same experiment steps are conducted in Jiading area. Three road icing prediction rules fitting Jiading district are got and the validation accuracy rate reaches 94. 87%. The above tests show that the road icing forecast model by C4. 5 decision tree algorithm can not only predict accurately and easily, but is facilitated direct- ly by forecasters. Meanwhile, each road icing forecast rule is in highly accordance with weather process and theory in meteorology.
出处
《气象科学》
北大核心
2015年第2期204-209,共6页
Journal of the Meteorological Sciences
基金
江苏省"青蓝工程"资助项目(2012)
