为了提高火灾事故预测的精度,根据我国火灾事故数据样本较小,波动性较大的特点,将遗传算法优化的灰色无偏预测模型与遗传算法优化的BP神经网络模型结合起来,建立灰色神经网络优化组合模型,充分发挥无偏灰色预测模型适用于小样本的数据预测的优势与BP神经网络处理非线性问题的优点。分别采用遗传算法优化后的无偏灰色GM(1,1)模型、遗传算法优化的BP神经网络预测模型与灰色神经网络优化组合模型对我国1998-2008年的火灾事故进行拟合,并对2009-2011年的火灾事故发生数进行预测。结果表明：灰色神经网络优化组合模型的预测误差最小,精度最高,适用于火灾事故的预测。%The prediction of fire accident is the basis for fire department planning and decision-making. The statis-tical data of fire accidents in China has the characteristics of small sample and big fluctuation. In order to improve the accuracy of fire accidents prediction, the model combining unbiased grey forecasting model and BP neural net-work method optimized by genetic algorithm was developed to adapt to the characteristics of fire accident statistical data. The model gave full play to the advantage of unbiased gray prediction model in fitting the small sample and superiority of BP neural network time series prediction for dealing with the non-linear problems. According to the statistical data of fire accident in 1998-2011, the optimized unbiased GM(1,1) prediction model, the optimized BP neural network time series prediction model and the optimized grey neural network combination model programed by Matlab were used to fit the number of fire accidents, and the 2012-2015 fire accident numbers were predicted. The results showed that the new model has fewer errors and better forecasting precision. Consequently, comparing with the traditional method, the new model is more applicable for the prediction of fire accidents.
展开▼
