基于SMOTE算法与决策树的沙尘暴短期预警研究

The Short-term Forecasting Model of Sandstorms Based on SMOTE and Decision Tree Learning Algorithm

针对沙尘暴灾害发生时间预测准确率较低、传统的预测模型预测效果欠佳问题,建立了基于SMOTE算法与决策树算法的沙尘暴预测模型.该模型利用西北六省的气象观测数据,较好地解决了稀有类的分类问题,总体预测成功率达到76.25%.研究结果表明该模型分类准确率高、泛化性能好、抗噪音、鲁棒性好,较好地解决了沙尘暴预测中不平衡样本的分类预测问题,可用于实际的沙尘暴预警.

As the traditional algorithms are defective in forecasting the accuracy of sandstorm disaster , this paper established a forecasting model combined SMOTE algorithm with decision tree learning algo‐rithm .Using meteorological observation data of six provinces in Northwest China ,the classification of rare class is well solved ,and the predictive accuracy rate reaches 76 .25% .The results showed that the model can be used for the actual sandstorm warning with good classification accuracy ,generalization performance , robustness and anti‐noise properties in solving the classification problems of the unbalanced samples in the sandstorm forecasting .