超啟發式多目標最佳化演算法於多準則存貨控制之研究

A Study of Multi-objective Meta-heuristic Algorithm for Multiple Criteria Inventory Control Problem

陳姿伶(Tzu-Ling Chen)

指導教授 ： 林建宏

義守大學/電機資訊學院/資訊管理學系/碩士(2014年)

https://doi.org/10.6343/ISU.2014.00453

摘要

存貨管理對企業來說是相當重要的一項管理工作，其目的是為了可以減少成本維持公司的服務水準，並降低缺貨的問題，以滿足顧客對商品的需求。當廠商同時訂貨的時候，安全庫存和服務水平問題也要明確的偏好銜接，而企業利用存貨控制來控管貨物，達到安全庫存和銷貨控管，降低缺貨的可能性。基本上，存貨管理為一個多準則決策問題。
本論文研究的目標是要設計、模擬、實作與驗證有效的超啟發式多目標最佳化搜尋演算法於多準則決策問題，並將實作應用於多準則存貨控制之問題。雖然多準則決策問題與解決最佳化問題的研究有相當多改善策略的發展，然而嘗試以多目標最佳化之計算智慧方法應用於多準則決策問題的研究卻不多。啟發式演算法是啟發構想於自然的學習方法應用於解決最佳化的問題上。而從實際的觀點，演化式演算法是一個群體式的、啟發式的最佳化搜尋方法。模仿一些自然演化的觀點與構想，於反覆執行過程中經由不斷選擇、更換去搜尋更適合的問題解決方案。本論文研究構想以設計多目標最佳化搜尋來達到區域開發與全域探索之演算法來解決多準則決策問題。此研究從分析到應用將試圖設計一個有效的啟發式多目標搜尋法於多準則決策上，從模擬到實作到驗證去瞭解如何與為何能有效的解決問題。實驗通常會伴隨著模擬來進行，本論文透過有效的解決多目標最佳化問題的模擬與實作進而轉換到實際案例。
本論文應用多目標蝙蝠演算法來研究缺貨後補的問題，希望成本降低但補貨的次數卻沒有因為太多而讓成本提高，能更有效的讓成本與次數能達到平衡，利用多目標得到最佳的解，讓缺貨後補的問題有得到有效的解決辦法。

關鍵字

存貨管理 ； 存貨控制 ； 多目標最佳化 ； 多準則決策 ； 啟發式演算法 ； 多目標蝙蝠演算法

並列摘要

The objectives of inventory management are to maintain high level of service quality by using least cost and to reduce the possibility of shortage in order to satisfy the requirements of customers at the meantime. The goals of this thesis are to design, model, simulate, implement, and verify efficient meta-heuristic multi-objective algorithms for the multiple decision making problems. Multiple criteria decision making research has become a main area of research for dealing with complex decision problems which require the consideration of multiple objectives or criteria. In Multi-objective optimization (MOO), solutions/alternatives are implicitly defined by a set of constraints that bound a feasible region, and objective functions are then optimized in this region (continuous problems) or large set of alternatives and their performances according to the multiple evaluation criteria is given explicitly (discrete combinatorial problems). Multiple criteria decision making (MCDM) addresses mainly discrete problems with not very large (combinatorial) sets of alternatives. Both MOO and MCDM involve evaluation of solutions/alternatives based on multiple criteria/objectives and search for trade-off solution(s) (e.g. using Pareto dominance). In some context, the benefits of utilizing MCDM include that conflicting design objectives are taken into account simultaneously leading to an overall insight of the problems which would deliver a significant and competitive advantage to the engineering design community. In some sense, the benefits of MOO include that the conflicting objectives are taken into account simultaneously, via practically implementing and testing Pareto-optimal solutions. It is very important that before the actual decision about the final solution takes place the decision maker should gain a good understanding about the trade-offs between the solution alternatives. Then the final decision can be firmly taken. Therefore, MOO approaches for creating Pareto-optimal solutions are considered vital to MCDM community. Implementing the MCDM task for solving optimization problem is considered as a very important and in the same time complicated approach for researchers to pursue. The aim of this thesis is to exploit synergistic meta-heuristics for specific topics and mathematical models in the field of multiple criteria decision analysis and multi-objective optimization and to algorithmic problems raised by the use of such models. Meta-heuristic algorithms are often nature-inspired, and they are becoming very powerful in solving optimization problems. To model the classic tradeoff between local search and global exploration, we investigate synergistic strategies for meta-heuristic multi-objective optimization learning, with an emphasis on the balance between intensification and diversification. To this aim, we propose a behavioral diversity preservation mechanism and multi-objective algorithm. This thesis intends to design a multi-objective bat algorithm, from analysis to their applications. We try to analyze the main components of this algorithm and how and why it work efficiently. The created algorithm is applied to multiple criteria inventory control problems.

並列關鍵字

inventory management ； inventory control ； multi-objective optimization ； multi-criteria decision making ； meta-heuristic ； multi-objective bat algorithm

參考文獻

9. 許顧輝(2005)。模擬退火法於設計繞射光學元件之研究。國立台北科技大學光電工程系碩士論文。

21. 許晉雄、鄒慶士(2010)。應用多目標為粒群演算法求解(s, Q)存貨控制最佳化模式。東吳經濟商學學報，第69期，47-82頁。

22. 許晉雄、鄒慶士(2010)。混和式多目標微粒群演算法在存貨控制模式之應用。Logistics Management Review，第四卷，第一期，23-44頁。

23. 許晉雄、鄒慶士(2010)。(S,Q)存貨控制缺貨後補模式之多目標最佳化分析。台灣管理學刊，第10卷，第一期，19-50頁。

24. 許晉雄、鄒慶士(2010)。多目標銷售損失存貨模式之決策問題。東海管理評論，第11卷，第一期，95-132頁。