摘要： 结合回归预测模型，提出一种对网络中单个组件故障率的多项式回归预测模型。通过将多项式回归模型转化成多元线性回归模型求解参数，得到预测方程。利用检验预测回归方程和回归系数显著性，得到同实际故障统计数据拟合的效果优于原方程的新预测方程。使用一组网络组件的故障率数据对预测方程进行验证，结果表明，在根据预测误差低于2%的要求而确定的预测区间内，该方法可以为网络中单个组件故障率预测提供决策依据。

关键词: 网络故障, 多项式回归, 多元线性回归, 故障预测, 故障率

Abstract: A single module failure rate prediction method is proposed using regression model. Through converting polynomial regression model to multiple linear regression to solve parameter value, the prediction equation is acquired. Checking significance of equation and coefficient, a new equation of which the fitting effects are better is established. The new equation is verified through some types of module failure rate data obtained from network experiment. The result shows that the proposed model can be used as a supporting model in the decision-making of a single module failure rate prediction.

Key words: network fault, polynomial regression, multiple linear regression, fault prediction, fault rate

中图分类号: 

• TP393