整合決策樹與多目標最佳化方法的參數決策過程

Abstract

在現代化的製程中，製程參數的設定通常會影響多個製程目標，因此如何設定合適的製程參數以達到決策者期望的製程目標成為一個具有挑戰性的問題。本研究提出一個數學規劃模型將多目標最佳化方法與決策樹模型相結合。我們以決策樹作為後設模型用於預測目標與製程參數之間的關係，並透過數學最佳化模型整合決策樹的預測結果和其他的製程要求，以提供適當的製程參數設定建議。透過案例研究，我們不僅證明了我們的模型具備有效性，並將其與基於多目標決策樹建構的方法進行比較。根據顯示的結果，我們提出的採用多個單一目標決策樹的方法所提供的解決方案的品質優於傳統的使用單一多目標決策樹的方法。此外，建議的參數設定為決策者提供了三種決策偏好選項，以協助他們根據製程目標偏好選擇最合適的參數設定。

This paper introduces an innovative approach to tackle the challenge of determining process parameter settings in order to achieve multiple process goals. We propose a mathematical programming approach that integrates a multi-objective formulation with decision tree models. The decision tree models predict different targets, and the mathematical model combines the outputs from the trees with other process requirements. The case study demonstrates the effectiveness of our model and compares its performance to a method based on multi-objective decision tree. The results reveal that our method, employing multiple single-objective decision trees, returns the superior solution than the conventional approach utilizing a single multi-objective decision tree. The suggested outcomes provide decision-makers with three preferences to assist them in selecting the most suitable option based on their target preferences.