考量個人化之多目標商品組合裝箱問題

Multi-objective Personalized Product Combination Optimization Considering Bin Packing

陳進彥(Jin-Yan Chen)

指導教授 ： 陳同孝 ； 陳民枝

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2019年)

摘要

在每逢節慶時，消費族群會對電子商務平台的零售業者，提出大量的商品組合需求。為避免提供過於制式化的商品組合，業者應提供個人化(Personalized)的服務，以滿足消費族群不同的偏好(Preference)。而該組合也需符合裝箱的條件，以便進行搬運。因此，本研究在消費族群的偏好劃分為四項：預算、商品種類的偏好、飲食習慣、不喜愛的廠牌，其中商品種類的偏好是以志願序方式填寫。在業者的組合限制中劃分為三項：重量、箱子容積、商品種類的限制。原先商品的三為問題簡化成一維，使用體積的方式進行求解。經以上方法透過基因演算法進行求解，並依商品組合的總價格和總偏好作為求解目標，最後挑選出符合消費族群偏好的商品組合。實驗結果顯示，加入消費族群的偏好和業者的組合限制，能夠求得出符合的商品組合；此外，為了驗證本研究方法在基因演算法中的可行性，則使用窮舉法求得出最佳解，並與基因演算法進行比較，得出結果與窮舉法的最佳解相同。而本研究也針對於基因演算法的參數值、挑選不同的品項數量、懲罰函數，以及價格、體積、偏好之權重值的調配進行多項實驗。結果證實本研究的方法，在透過基因演算法中可求得出符合消費族群偏好的商品組合，並且該結果也符合預算、重量以及裝箱條件。在結果中也與莊雅筑進行比較，在平均總體適應函數中獲得1.05171，而莊雅筑為3.60493，相較之下本研究結果較好。在多項的商品資料中，透過此方式可以大量節省業者挑選商品的時間，以找到符合消費族群和業者對組合的條件之商品。而挑選出的商品組合也不會過重，造成搬運時的傷害。

關鍵字

基因演算法 ； 商品組合 ； 最佳化 ； 個人化 ； 零售業 ； 多目標 ； 裝箱問題

並列摘要

There are large demands for packaged products offered retailers consumers on the e-commerce platform to consumers during festive seasons. In order to avoid offering basic standardized product combinations, retailers should provide personalized services to meet the different preferences of the consumers. The product combinations also need to meet the packing conditions for handling. Therefore, the preferences of the study in for the consumer are divided into four categories: budget, product category preferences, eating habits, and brand preference. The preference for the product category is filled in by preference order. The retailer's combination limit is divided into three categories: weight, bin volume, and product category restrictions. The three dimension of the original product are simplified into one dimension, and the volume is used to solve the problem. Results showed that by adding the preferences and packing conditions of the consumer group, it is possible to find a product combination that meets the consumer requirement. In addition, the feasibility of this research method in the genetic algorithm is verified by using the exhaustive method to obtain the best solution. Both results are compared which showed the genetic algorithm results are the same as the best solution of the exhaustive method. This study also conducts multiple experiments on the parameter values of genetic algorithms, namely, the selection of different item quantities, the penalty function, and the assignment of weight values. The results showed that the research method can find a suitable product combination in the genetic algorithm. The results also meet budget, weight and packing conditions. In the results, compared with Ya-Chu Chuang, the research method obtained 1.05171 in the average total fitness function, and Ya-Chu Chuang built 3.60493. After the comparison, the results of this study are better. In a lot of good data, this way can save a lot of time for the industry to select goods, in order to find the goods that meet the conditions of the consumers and the retailers. The selected combination of goods will not be too heavy, causing physical harm during handling.

並列關鍵字

Genetic algorithm ； product combination ； optimization ； personalization ； retail ； multi-objective ； bin packing problem

參考文獻

[1] M. Beyaz, T. Dokeroglu and A. Cosar, “Robust hyper-heuristic algorithms for the offline oriented/non-oriented 2D bin packing problems,” Applied Soft Computing, 36, p. 236-245, 2015.

[2] M. Buljubasic and M. Vasquez, “Consistent neighborhood search for one-dimensional bin packing and two-dimensional vector packing,” Computers & Operations Research, 76, p. 12-21, 2016.

[3] E. Cao and M. Lai, “An improved differential evolution algorithm for the vehicle routing problem with simultaneous delivery and pick-up service,” Third International Conference on Natural Computation, 3, p. 436-440, 2007.

[4] A. E. Eiben and S. K. Smit, “Evolutionary algorithm parameters and methods to tune them,” Berlin: Springer, p. 15-36, 2011.

[5] A. Grange, I. Kacem and S. Martin, “Algorithms for the bin packing problem with overlapping items,” Computers & Industrial Engineering, 115, p. 331-341, 2018.