圖書採購問題最佳化之研究

Abstract

採購圖書資料時，採購人員普遍面臨物價膨脹、預算縮減的壓力。如何在眾多圖書資料中利用有限的預算，選購符合圖書館館藏方向，且滿足讀者的需求，是圖書館採購人員經常面臨的問題之一。為解決此問題，本文旨在提出圖書採購問題最佳化的架構，針對四個實務問題進行研究，分別以整數規劃方法建立四個採購問題的模式，設計求解該問題之離散粒子群最佳化演算法(discrete particle swarm optimization, DPSO)，透過電腦模擬實驗驗證演算法的效能與效率，藉此證實所提架構之可行性。 首先，本論文探討在集中式預算下，考量各種類別圖書資料採購冊數上下限的限制，挑選哪些圖書資料，可最大化讀者對於所採購圖書資料的平均喜好度。其次，本研究探討在分散式預算下，考量各種類別圖書資料採購冊數上下限，及各種語言別圖書資料採購冊數上下限的限制，選購哪些圖書資料，可最大化讀者對於所採購圖書資料的總喜好度。再者，本研究探討在分散式預算下，考量前述問題相同的限制，選購哪些圖書資料，且費用由部門或系所共同分攤，可最大化讀者對於所採購圖書資料的平均喜好度及預算執行率。最後，本研究探討在集中式預算下，考量各種類別圖書資料的採購金額比重上下限、各種語言的採購金額比重上下限、各種採購方式的採購金額比重上下限等，該選購哪些圖書資料、該與何廠商用何種方式採購，可最大化讀者對於所採購圖書資料的平均喜好度。

The price inflation of library materials, the shrinking of library budget, and the growth of electronic resources continue to challenge library materials acquisition. Subject to the requirements of various fields of patrons, one of the most challenging issues is to acquire materials fairly, and to ensure that the acquired materials attain the highest and best use of the budget. This study proposes an optimization framework of the library materials acquisition problems. To demonstrate the applicability of the proposed framework, four variants are formulated in integer programs and tailored discrete particle swarm optimization (DPSO) is deployed to produce approximate solutions. The first variant, Average Preference Maximization Problem with Centralized Budget (APMP with CB), is to maximize the average preference of the acquired materials. The decisions are to determine which materials should be acquired under the constraints of centralized budget and the limit on the number of materials in each category. To demonstrate the feasibility and applicability of the proposed DPSO algorithms, computational experiments are conducted. Computational results show that the proposed approaches are able to provide quality solutions for the problem in assorted scenarios within a reasonable time. The second variant, Total Preference Maximization Problem with Decentralized Budget (TPMP with DB), is to maximize the total preference of the acquired materials. The decisions are to determine which materials should be acquired by which departments under the constraints of departments’ budgets and the limit on the number of the acquired materials in each written language and in each category. Two different constraint-handling mechanisms are designed for the applied DPSO algorithm. It is evident from the computational results that one constraint–handling mechanism can solve the problem effectively and efficiently, while the repair operator takes more execution time. With the same decision and constraints as the second variant, the third variant, Average Preference and Execution Rate Maximization Problem with Decentralized Budget (APERMP with DB), is to maximize the average preference and execution rate of the acquired materials. The decisions are to determine which materials should be acquired and which departments should cover the cost associated with those materials under the constraints of departments’ budgets and the limit on the number of the acquired materials in each written language and in each category. To tackle the constrained problem, a DPSO with scout particles is presented. A series of computational experiments are designed and conducted. The results are statistically analysed, and it is evinced that the proposed DPSO is an effective approach for the studied problem. The fourth variant, Total Preference Maximization Problem with Centralized Budget (TPMP with CB), is to maximize the total preference of acquired materials. The decisions are to determine which materials should be acquired from which vendor by which acquisition method under the constraints of centralized budget and the limit on the cost of the acquired materials of titles, packages, acquisition methods, each written language, and each category.