基于随机森林的统计控制图模式识别研究

引入随机森林方法进行统计控制图模式识别的研究。提取了控制图的统计特征和形状特征，设计了5种不同的特征组合方法，利用蒙特卡洛仿真方法产生训练数据集和测试数据集，选取了常用的3种模式识别方法（支持向量机方法、人工神经网络方法、决策树方法）进行对比。实验结果表明，随机森林方法相比其他3种分类器方法，在分类准确率和消耗时间两个维度上都有明显优势，可以应用于统计过程控制图模式识别。

A Study of Pattern Recognition of Statistical Control Chart Based on Random Forest

The random forest method is introduced to study the pattern recognition of statistical control chart. The statistical features and shape features of the control graph were extracted, five different feature combination methods were designed, and Monte-carlo simulation method was used to generate training data set and test data set. Three commonly used pattern recognition methods (support vector machine method, artificial neural network method and decision tree method) were selected for comparison. Experimental results show that compared with the other three classifier methods, the random forest method has obvious advantages in classification accuracy and consumption time, and can be applied to the pattern recognition of statistical process control chart.