油料消耗神经网络组合预测模型

Research on the Combination Forecasting Model of Neural Network for POL Consumption

油料消耗单一预测模型精度不高，难以适应信息化条件下精确保障需要。以单一的神经网络预测模型、时间序列预测模型和灰色预测模型为组合预测的基础，利用神经网络求取3种预测模型的组合预测权重系数，将这3种单一预测模型的预测结果作为神经网络组合预测模型的输入，求得一个新的预测结果。平均相对误差和均方差比表明，神经网络组合预测模型比单一预测模型更为优越。

Single forecasting model for POL consumption is not precise enough and can＇ t meet the needs of precise support under the condition of information warfare. Based on the combination of single neural network forecasting model, time sequence forecasting model and gray forecasting model, neural network is used to resolve the weight coefficient of the combination for each forecast model. Forecast results from each single forecasting model are input to neural network forecasting model to get new forecast results. Finally, according to average relative error and average variance ratio,the combination forecasting model of neutral network is proved to be superior to the single forecasting model.