以基因演算法求解競爭性流量截取區位設置問題

Abstract

在現今的社會，許多人的購物習慣通常是順道的形式而非專程購物，尤其是下班時段常有大量餐飲、購物的順道旅次，但是多數區位設置的研究仍是考量專程購物旅次之「涵蓋人口最大化」。而非順道購物旅次之「截取顧客最大化」。此外先前多數研究僅就目標地區進行設施區位選擇，並未明確指出該地區是否存有相同性質的設施，在已有現有設施的狀況下，新的競爭者如何加入目標地區來搶奪市占率？因此本研究提出在競爭的環境下，如何設置新的設施才能最大化得以被新設施截取的流量，使得企業在競爭當中設站時有新的方法能考慮。 本研究的假設是在一特定區域市場中，已存有競爭對手所設立之商店，我方也欲在此區域設置商店。本研究之研究模式是由競爭型流量截取設置模式發展而來，已知需求會透過網路空間經濟地移動到設施接受服務，透過理想的設施設址，爭取截取(Capturing)到之顧客最大化。針對此問題，本研究根據文獻提出的模式，選擇使用基因演算法的概念，以發展出更好的求解方式來解決此問題。 有關數值驗證部分，本研究以二十五點中型路網來進行數值測試，並考量路網內不同數目之新設施，以研究新競爭者設施數量與求解效果的關係。數值測試中既存商店之位置以及需求流量以亂數產生，實驗的結果顯示，透過基因演算法求解結果優於文獻中貪婪法(Greedy Heuristics)的求解品質，而求解時間也沒有相差太多。次外，把問題加大探討問題規模對求解的影響，發現例題越大基因演算法的求解優於貪婪法越多，驗證基因演算法對於大問題更具求解上的優勢。本研究的成果，應該可以做為店家考慮流量需求時設點決策上之參考依據。

Nowadays, more and more people go shopping on the way home or to the working place, rather than take a special trip for shopping. However, most of studies focus on maximizing the “covered” potential shoppers (making a dedicated trip), not maximizing the “captured” business flow (based on the pass-by trips). In addition, most of the previous studies aim to select the facility nodes from the potential facility site, without considering whether there are some existing facilities in the region. How would the newly introduced facilities affect the market share? Thus, the objective of this study is to efficiently determine the location of the facilities to maximize the captured customer flow in a competitive environment. Given a specific market region, it is assumed there are some existing retailing facilities set up by the competitor, and the new operator plans to set up its own facilities to maximize its market share. From the academic point of view, this study is based on the competitive flow-capturing location allocation problem (FCLAP) and assumes that the potential customers can detour, away from their pre-determined route, to receive the service. The objective is to capture the most customers by the ideal facility allocation. To this problem, the model proposed in a past study is used. In addition, the Genetic Algorithm is used to design a solution algorithm that can serve as a better method to handle this problem. In the numerical experiment, the proposed model and the solution algorithm have been tested with a middle-sized example network with 25 nodes respectively. The influence for the number of new facilities in the network has been examined. In designing the test problems, we simulate the locations of the existing facilities and the demand flows by random numbers. It is found the solution algorithm based on the genetic algorithm can achieve a better solution quality when compared with the greedy heuristic algorithm in the literature. Furthermore, the example network has been extended to examine the effect of problem scale. It is found that the genetic algorithm does perform better than the greedy heuristic algorithm with bigger networks. The test results show the developed model and solution approach can be used for dealing with the Flow Capturing Location-Allocation Problem and providing the decision support for the planners.