基于回归分析的雷灾死亡人数预测模型

利用1995—2016年广东省雷电灾害伤亡数据,对广东省雷电灾害中因雷击引起死亡的人数进行了统计分析。结果表明:1995—1998年为雷灾死亡人数的上升期,1998—2016年总体呈下降趋势,其中2004年雷灾死亡人数突增出现奇点。2008年前后雷灾死亡人数变化明显,1995—2008年每年雷灾死亡人数波动较大,年平均死亡人数为66人;2009年后雷灾死亡人数波动平缓,年平均死亡人数为12人,同比减少了81.8%。对比分析五类拟合曲线后,得到广东省1998—2016年雷灾死亡人数年际动态变化预测模型的最佳拟合曲线是负指数曲线。为减少数据样本波动大和瞬间毛刺多的问题,通过百分位分析法,得到了2017年和2018年的预测值分别为11和10。t分布假设检验给出了2017年和2018年置信度为95%的预测区间上下限分别为24、6和25、6。

The forecasting model of lightning fatalities based on regression analysis

Based on 1995-2016 Guangdong province lightning disaster cases collected by Guangdong lightning prevention and disaster reduction management center, a statistical analysis of lightning striking fatalities was carried out. The results indicated that the number of deaths rose from 1995 to 1998 while declined during 1998 and 2016, and presented a sudden increase in 2004. The number of deaths due to thunderstorms changed significantly around 2008. From 1995 to 2008, the annual number of deaths caused by thunderstorms fluctuated greatly, with the average number of 66 deaths per year. After 2009, the annual number of deaths tended to fluctuate gently, with an annual average of 12 deaths per year, and the year-on-year decrease was 81.8%. Comparing the five types of fitting curves, the best fitting for the predicted inter-annual variation of lightning striking fatalities in Guangdong province from 1998 to 2016 was a negative exponential curve. In order to reduce the problem caused by data fluctuations and excessive transient peaks, the predicted values of 2017 and 2018 were 11 and 10 by using percentile analysis, respectively, and the prediction confidence interval of 95% confidence given by t-distribution hypothesis test was 6-24 in 2017 and 6-25 in 2018.