以子集合求解大規模方向距離函數及加法模式問題

Abstract

有鑒於大規模線性規劃問題產生的計算負擔，本論文提出一個利用子集合求解大規模問題之演算法，針對資料包絡分析之方向距離函數及加法模式問題求解，不同於一般演算法找尋問題的近似解，本論文提出的演算法保證解的收斂，利用子集合求解大規模問題讓限制可使用變數的試用版求解軟體得以求解大規模的問題，並在大部分的案例分析中縮短部分收斂所需的時間，增加求解的計算效率。

This thesis investigates how to solve large-scale problems in data envelopment analysis (DEA) by using small data sets. In particular, we focus on directional distance functions and additive models. The thesis guarantees the optimality of solutions in both models and saves the computation time. Using small data sets to solve the problems, the thesis makes the trial version software available to solve the large-scale problems.