随着神经网络在实际生产中日益广泛的应用，有必要对网络模型输出结果的精确度进行估计。本文介绍了一种计算置信区间的方法，推导出目前广泛应用的ＢＰ网的置信区间计算公式。针对目前计算置信区间过程中存在的一些问题，提出了对置信区间长度的统计修正方法，使其用于工厂在线预测中的可靠程度得到提高 With the increasingly wide application of neural network in practical production, it is necessary to estimate the accuracy of the output of the network model. This paper presents a method to calculate the confidence interval and deduces the formula for calculating the confidence interval of BP network which is widely used at present. Aiming at some problems existing in the calculation of confidence interval, a statistical correction method for the confidence interval length is proposed, which improves the reliability of the method for on-line forecasting in factories