在化學的領域裡，不斷的實驗、合成、量測，是探討議題的最基礎方法；在統計學的角度，資料探勘亦是長久發展且已成熟的方法，許多行銷公司將其用在消費者行為討論。本文將資料探勘的想法應用於自然科學領域，以區別分析和非線性迴歸減少化學實驗的實際操作，減少計算值與實際值的誤差，並且對於化學品的信賴性做出兩群組的預測，改善可以通過信賴性實驗的機率。 以統計學針對過去資料有效的分析，在實驗上是助益頗大的，不僅可以減少試誤的實驗數，亦可提供明確的方向給予設計者作為參考。 In this paper, the ideal is analysis customized production in the statistical point of view the point and improves the reliability. Nonlinear Estimation Procedures is to reduce the difference between real value and calculated value. And classification from Categorical is used to clear two groups, one is reliability pass, another is failure, and predict the production reliability. The major benefit is decreasing the try and error experiments and increasing the probability of reliability pass.
