投影追踪回归(PPR)法建立卷烟焦油预测数学模型

Establishment of Mathematical Model for Predicting Cigarette's Tar Yield with Projection Pursuit Regression Method

为了有效、准确地控制成型纸、接装纸和卷烟纸(简称"三纸")对卷烟焦油量的影响,利用投影追踪回归(PPR)的方法,建立了通过"三纸"预测卷烟的焦油数学模型,并与传统的多元参数回归模型进行了比较。结果表明:两种模型对单排激光打孔和双排激光打孔卷烟的焦油进行预测,其还原检验准确率分别为88%、80%和61%、59%,预留检验准确率分别为60%、75%和40%、50%,与传统的多元参数回归模型相比,PPR所建模型准确度较高。

In order to control the effects of parameters of plug wrapper, tipping paper and cigarette paper on cigarette＇ s tar yield effectively and accurately, a mathematical predicting model with Projection Pursuit Regression （PPR） was established and compared with the traditional multiple parameter regression model. The two models mentioned above were used to predict the tar deliveries of single- and double-line laser perforated filter cigarettes, their corrected accuracies were 88% , 80% and 61% , 59% , respectively; and their accuracy against reservative control were 60% , 75% and 40% , 50% , respectively. The PPR model is more accurate than the traditional one.