基于隶属度模糊最小二乘支持向量机的工序能力预测

Intelligent Prediction for Process Capability Based on FLS-SVM

提出了一种基于隶属度模糊最小二乘支持向量机(FLS-SVM)的时间序列预测新方法。一方面,该方法较好地解决了小样本学习问题,避免了人工神经网等智能方法在对小批量生产工序能力进行预测时所表现出来的过学习、泛化能力弱等缺点;另一方面,由于对于历史数据实行的重近轻远的原则,使得该方法预测精度高且容易实现。实验表明,该方法具有很好的有效性与实用性。

A new time series prediction method was put forward,based on fuzzy least square support vector machine（FLS-SVM）.On the one hand,it can solve the small-batch learning better and avoid the disadvantages,such as over-training,weak normalization capability,ect.,of artificial neural networks prediction.On the other hand,this proposed method is more accurate and can be realized easily because it sets larger weight on the nearer sample but smaller weight on the farther.As a result,the prediction effect shows the effectiveness and practicability of the mode.