基于BP神经网络的建筑施工事故非线性组合预测
Nonlinear combined forecasting of construction accident based on BP neural network

以建筑行业百万元产值死亡率为预测对象，建立了非线性回归模型、三次指数平滑模型、灰色模型、线性组合预测模型和基于BP神经网络的非线性组合预测模型. 结果表明，该非线性组合预测模型的拟合及预测精度均较其它模型有明显提高，能够有效地综合利用各单项预测模型所提供的信息，证明了该模型适用于对建筑施工事故的宏观预测，为非线性组合预测模型的构建提供新的思路. 应用该模型对2014-2016年全国建筑施工百万元死亡率进行预测，计算结果表明，未来几年建筑业安全生产将会保持在一个较稳定的水平.
The nonlinear regression model，the cubic exponent smooth model，the grey model，the linear combined and the nonlinear combined forecasting model based on BP neural network were established for the forecasting of the mortality of per million output value of construction industry in China（MMCC）. The compare of different model’s results showed that the precision of fitting and forecasting of nonlinear combined model were obviously higher than the other’s，which means its ability of integrated utilization of effective information provided by each single model，showed that the model is suitable for the macro-forecast of the construction accident，and develop a new way to establishing the nonlinear combined prediction model. The MMCC from 2014 to 2016 were forecasted in the use of the model，results showed that the safety of construction in the next few years will remain relatively stable level