概率天气预报的K近邻非参数估计仿真模型

基于模式识别和相似预报思想，提出了一种制作概率天气预报的K近邻非参数估计仿真模型(简称KNN-M)。该模型包括历史样本数据库、近邻子集搜索程序、近邻子集优化算法和预报量估计技术。利用该模型进行了降水和云量的概率预报试验，独立样本检验结果表明，该模型预报稳定性好，预报准确率较高，具有较好的业务应用前景。

K-Nearest Neighbor Nonparametric Estimation Bootstrap Model for Weather Probability Forecasting

Based on the model identification and analogue forecasting, a new approach for probability weather forecasting is proposed, which is called K-nearest neighbor nonparametric estimation bootstrap model (named KNN-M for short). This model includes a historical database, a procedure of searching for the nearest neighbor subset and its optimization algorithm and the technique of predict and estimation. The case experiments are made on the forecasting of precipitation and cloud amount by using KNN-M. The individual sample test results show that the forecast estimation is stable and the accuracy is high. KNN-M has good prospects in operational weather forecast.