Research On Competitive Online Ordering Strategy By Aggregating Expert Advice With Periodic Demand

As a classic problem of inventory management research,the newsvendor problem is often widely used in many fields of economic life.In traditional research on the newsvendor problem,statistical assumptions about demand are usually made and models are solved on this basis.However,in the actual decision-making environment,it is often difficult for retailers to obtain complete information about the distribution of demand,and market demand is not always can carry on accurate portrayal in a certain probability distribution.The increasing maturity of machine learning techniques has provided new ideas for interdisciplinary integration research,and many online algorithms have been introduced into the field of economic management research to help solve practical management decision-making problems.The WAA algorithm is an online algorithm emerging from the discipline of computational science.It gets rid of the statistical assumption of input data and only makes sequential decisions based on progressively updated historical data.Therefore,it can be used to solve the newsvendor problem with unknown demand distribution.In this thesis,we apply the WAA algorithm to aggregate static expert advice to study the multi-period newsvendor problem without statistical assumptions,and give an online ordering strategy for the extension newsvendor model.First,the continuous and discrete multi-period newsboy decision problems are explored separately by combining the periodic demand characteristics,and specific online ordering strategies are given.The competitive performance of the strategies is analyzed theoretically,and then the revenue performance and competitive performance of the strategies are tested with numerical arithmetic examples to check the effectiveness and robustness of the strategy.Secondly,considering both over-ordering and under-ordering,the model is extended by introducing repurchase value and out-of-stock loss to investigate the effects of both on newsvendor decisions,and online ordering strategy under continuous and discrete demand are given.Finally,in the context of the prevalence of e-commerce,the conditional free shipping policy is considered to do an extended study on the newsvendor problem,combining shipping cost and free shipping threshold to develop an online ordering strategy for newsvendor,and the effect of different values of the parameters on the performance of strategy revenue and competitive performance is further tested in the numerical algorithm.The online strategies designed in this thesis are free from the demand assumption statistics and give theoretical proofs of the competitiveness of the strategies.The results of numerical arithmetic examples also further verify that the online ordering strategies constructed by applying the WAA algorithm have good revenue performance and competitive performance,and the cumulative revenue achieved by the online strategies can be comparable to that achieved by the optimal expert as the number of decision periods increases.This thesis combines realistic factors such as periodic demand,repurchase value,out-of-stock loss and conditional free shipping policy to construct the newsvendor model and give the online ordering strategy.On the one hand,it enriches the related research on inventory management of perishable goods,and on the other hand,it provides a theoretical reference for ordering decisions of perishable goods retailers.