基于Logistic回归、Markov过程和改进灰自助法的导弹备件需求预测

 为了提高预测间断性需求导弹备件的精度,提出一种基于Logistic回归、Markov过程和改进灰自助法的组合预测模型。将样本序列划分为解释变量序列和自相关序列,对解释变量采用Logistic回归模型预测提前期非零需求发生概率,对自相关序列采用Markov过程估计提前期非零需求发生概率,把这两方面组合得到提前期非零需求发生概率,再运用改进灰自助法进行需求分布确定,得到最终的提前期需求。改进灰自助法先进行Bootstrap抽样,进行GM(1,1)二次数据拟合,既克服了Bootstrap法在小子样下的重复抽样问题,又克服,Bootstrap法在小子样下仿真结果不可信的问题。实例表明,提出的组合预测方法降低了预测误差,说明了该方法的有效性、可行性和实用性。

Demand Forecasting of Missile Spare Parts Based on Logistic Regression,Markov Process and Improved Grey Bootstrap Method

In order to enhance the forecasting accuracy of intermittent demands of missile spare parts, a combined forecasting model based on the logistic regression,Markov process and the improved improved grey Bootstrap method is proposed. This model splits the sample series into the explanatory series and the auto-correlated series. The probabilities of nonzero demands for the explanatory series in the lead time is estimated by a Logistic regression model, the probabilities of nonzero demands for the auto-correlated series in the lead time is estimated by the Markov process, and they are combined to obtain the probabilities of nonzero demands in the lead time. Finally the demand distribution is determinated by the improved grey Bootstrap method, where, the bootstrap sampling is made, and the data are matched by GM(1,1). Based on the principle of the grey bootstrap, the resample method is improved to avoid the bootstrap being repeatly resampled in a small sample case, and the GM(1,1) twice data fitting model is used to solve the problem of the credibility of the bootstrap's simulated result in a small sample case. Experimental results show that the combined forecasting model can significantly reduce the prediction errors and the method is effective, feasible and practical for forecasting the demands of missile spare parts.