基于数据挖掘降雨量建模和预测

ON RAINFALL MODELLING AND PREDICTION METHOD BASED ON DATA MINING

降雨量的大小会严重影响到一个地区水的质量。基于一个地区的雷达反射率数据和翻斗式雨量计(TB)数据,采用数据挖掘的方法进行降雨量的建模和预测。结合基于TB和基于雷达的降雨量预测模型的优点,提出一种充分利用TB数据和雷达数据进行降雨量预测的新模型。在这种预测模型中采用五种数据挖掘的方法:神经网络、随机森林、分类和回归树、支持向量机和K-最近领域法。为了分析模型的准确性和稳健性,以一种基于历史数据的基准模型和一种基于临近区域TB数据的模型用于对比。通过与几种模型的比较验证了该模型的准确性和有效性。

The size of rainfall may severely affect the water quality in a region. In this paper we use data mining approach to model and predict the rainfall based on radar reflectivity data and tipping-bucket （TB） data in a region. By combining the advantages of TB-based and radar-based rainfall prediction models, in the paper we resent a new model which makes full use of TB data and radar reflectivity data to predict rainfall. In such prediction model five data mining algorithms are used： the neural network, the random forest, the classification and regression tree, the support vector machine, and the k-nearest neighbour. To demonstrate the accuracy and robustness of the proposed model, we also consider a historical data-based benchmark model and a neighbouring region TB data-based model for comparison. The accuracy and effectiveness of the proposed model are shown by the comparison between a couple of models.