数据时间维度对火灾形势预测精度的影响

Influence of Data Time Dimension on the Accuracy of Fire Situation Prediction

为提高火灾形势预测精度,基于我国2008—2019年以年、季和月为时间维度的9项火灾历史数据,利用主成分分析法得到相应的火灾形势评价结果,根据评价结果并控制数据量构建GM(1,1)预测模型,通过比较后验差比值、小误差概率和平均相对误差,研究不同时间维度的火灾历史数据对火灾形势预测精度的影响,从而以适合的数据构建更加精准的预测模型。结果表明:(1)在预测2020年火灾形势发展趋势时,火灾历史数据的时间维度对GM(1,1)模型的预测精度有着较大的影响,季度数据尤其明显,且当数据量为3时,季度数据所建立的GM(1,1)模型的预测精度最高,为99.98%;当数据量为4~8时,年份数据所建立的GM(1,1)模型的预测精度普遍优于季度和月份数据,预测精度最高为96.74%(数据量为8);当数据量为9~12时,月份数据所建立的GM(1,1)模型的预测精度表现最高,最高为95.53%(数据量为12);(2)预测得到2020年全年、第1季度和1月份的火灾形势综合评价得分分别为0.166 0、0.949 1和0.733 5。可见,利用季度和月份数据可以细化和完善消防规划,从而更有效地降低火灾发生次数及其损失。

In order to improve the accuracy of fire situation prediction,based on nine fire historical data in China from 20o8 to 2019 with annual,quarter and monthly dimensions,the corresponding fire situation e-valuation results are obtained by using principal component analysis,and a GM(1,1)model is constructed according to the evaluation results and the amount of data is controlled,and the impact of historical data of different time intervals on its prediction accuracy is studied by comparing the posterior difference ratio,small error probability and average relative error,so as to build a more accurate prediction model with suit-able data.The results are as follows:OIn predicting the development trend of the fire situation in 2020,the time interval of historical data has a greater impact on the prediction accuracy of the GM(1,1)model,espe-cially the quarterly data.When the data volume is 3,the prediction accuracy of the model built by the quar-terly data is the highest,which is 99.98%;when the data volume is 4-8,the accuracy of the model built by the year data is generally better than that of the quarter and month,and the highest is 96.74%(the data volume is 8);when the data volume is 9-12,the monthly data performance is the best,up to 95.53%(the data volume is 12).@The fire situation for the full year,first quarter and first month of 2020 is forecast to be 0.1660,0.9491 and 0.7335.Quarterly and monthly data can be used to refine fire planning,thereby more effectively reducing the number of fire and the loss.