基于加权线性回归模型组的湖北省工矿事故死亡人数分析预测  

Analysis and forecast of death toll for industrial and mining accidents in Hubei province based on weighted linear regression model group

作　　者：彭东[1] 罗周全[1] 秦亚光[1] 王婷玉[1] 

机构地区：[1]中南大学资源与安全工程学院,湖南长沙410083

出　　处：《中国安全生产科学技术》2015年第11期167-173,共7页Journal of Safety Science and Technology

基　　金：国家自然科学基金项目(51274250)

摘　　要：以2005-2014年湖北省工矿企业事故死亡人数为基础,通过对时间序列图分析可知,该省工矿企业事故具有明显的周期性、季节性。另外,依据安全系统特征属性,系统近期的状况相对远期的状况对未来预测更具有影响力。因此,各个季节数据的权重应有所不同。综合考虑以上两种情况,提出对4个季节建立不同的线性加权线性回归模型,并进行组合,形成加权线性回归模型组,进而对2015-2016年各个季度进行预测。最后,与传统的时间序列分析方法比较可知,加权线性回归模型组的方法建立的模型显著性、拟合度更高,并且模型的可读性更强、更简单。Based on the death toll of the industry and mining accidents in Hubei province from 2005 to 2014, the time series diagram was analyzed, and the results showed that the accidents had obvious periodicity and seasonality. In addition, according to the characteristic attribute of safety system, the influence of recent status on future fore- cast is larger than that of long-term status, so the weights of data in each season should be different. Considering the above two situations, it was put forward to establish different weighted linear regression models for four seasons, and form the weighted linear regression model group by carrying out combination. Then the forecast was conducted on each season of 2015 and 2016. It showed that compared with the traditional time series analysis methods, the significance and fitting degree of model established by weighted linear regression model group method are higher, the readability is stronger, and the model is more simple.

关 键 词：加权线性回归 模型组 工矿事故 时间序列 预测 

分 类 号：X928.03[环境科学与工程—安全科学]