摘要
针对传统的分类方法在构建降雨预测模型时都存在着泛化能力低、精度不足的问题,基于集成学习的思想,提出一种改进的Adaboost-C4.5算法。通过自适应增强算法集成C4.5决策树算法,得到多个弱分类器,再赋予弱分类器权值,利用粒子群算法对其权重系数进行优化,最后线性组合成强分类器来提高模型的分类性能。文中选取大气压强、气温、风向、风速和相对湿度作为预报因子,并建立5个等级预报降雨模型。实验表明,所提模型在性能上表现更好,提高了预报的准确率,降低了预报的漏报率,在5个等级预报中,降低了3级和4级降雨预测的标准误差。
In view of the problems of low generalization ability and low accuracy of traditional classification methods in constructing rainfall prediction model,an improved Adaboost-C4.5 algorithm is proposed based on the idea of integrated learning.The C4.5 decision-making tree algorithm integrated with adaptive enhancement algorithm is adopted to obtain multiple weak classifiers.After the weight is endowed to the weak classifiers,the particle swarm optimization algorithm is used to optimize their weight coefficients.A strong classifier is linearly combined to improve the classification performance of the model.In this paper,atmospheric pressure,air temperature,wind direction,wind speed and relative humidity are selected as the prediction factors,and five levels of rainfall prediction models are established.The experimental results show that the proposed model has better performance,which has improved the accuracy of forecast and reduced the false alarm rate.
作者
胡玉杰
杜景林
董亚
滕达
HU Yujie;DU Jinglin;DONG Ya;TENG Da(Nanjing University of Information Science&Technology,Nanjing 210044,China)
出处
《现代电子技术》
2021年第14期6-10,共5页
Modern Electronics Technique
基金
国家自然科学基金项目(41575155)。
