关键词: 华南暴雨日数, 特征提取, 调整$Ps$评分, 最小角回归

Abstract: In this paper, we study the relationship between the rainstorm days in South China and the early atmospheric circulation factors as well as the external forcing factors, and based on it , build a model to forecast the rainstorm days in the area. Firstly, we select the high correlation regions to construct the prediction characteristics for variable dimension reduction. Then, a least angle regression model is built to forecast the rainstorm days in early summer in South China. Compared to other models in terms of time anomaly correlation coefficient ($TCC$), rate of the same sign ($SS$), coefficient of determination ($CD$), and adjusted prospect anomaly synthetical score ($APs$), it is obtained that the prediction results of the minimum angle regression model based on the high correlation region have a strong time correlation with the observed values and the high $APs$ score, which shows that the method constructed in this paper has a strong practical value.

Key words: Rainstorm days in Southern China, Feature selection, Adjusted $Ps$-score, Least Angle Regression