Two algorithms for classification on quality control

Abstract

In this article, no probability density and no prior probability are assumed for the good and defective classes, except possibly two means. A sequence of unidentified input lifetimes is used to estimate the percentages of two classes. After the classifier system is put into use, the new coming unidentified input lifetime can improve the percentages. A Monte Carlo simulation study with exponential or Weibull or gamma distribution on life is presented to demonstrate the classification. The study can be widely used in quality control.

本文先假設好壞兩條生產線的壽命分布是未知 ， 平均壽命低為不良品之生產線 ， 在僅知好壞兩條生產線之平均壽命下 ， 利用一組未分類的壽命樣品之資料 ， 去估計好壞兩類產品的百分比 ， 分類系統是利用一新加入未分類樣品壽命觀測值與兩平均壽命遠近來作產品分類並改進不良率估計。 蒙地卡羅數值研究 ， 是假設壽命分布來自指數、韋伯或Gamma分布下 ， 探討所提出兩種程序演算法之準確度及作未分類產品之品質分級。

(*聯絡人: [email protected])

Keywords:

• Bayes decision rule
• quality control
• stochastic approximation
• unsupervised learning
• climbing algorithm

Keywords:

• 關鍵詞： 貝氏決策法則 ， 品質管制 ， 隨機近似法 ， 未督導學習法 ， 爬坡法

Acknowledgment

The author is grateful to the editor and referees for valuable suggestions for the improvement of this article.

Notes

(*聯絡人: [email protected])