Uncertainty Oriented Collaborative Demand Forecast In Supply Chain

Nowadays, demand visibility of supply chain management has become an impor-tant issue. Demand-pull supply chain is more accurate than sale-push way, which hasbetter adaptability to meet customers'demand and obtains better returns. However, asthe supply chain environment becomes more uncertain and its complex structure, so leadto difficulties in obtaining accurate business requirements. As a famous phenomenon,bullwhip effect can cause widespread waste of resources of business and society. Howto get high-quality demand forecast has become a big problem that companies must face.The enterprises can select the following strategy. First, they can join the supply chain andshare information, resources, and forecasting with upstream and downstream enterprises.They can also reduce the uncertainty of demand forecasts to improve demand forecastingresults. Secondly, the enterprises can strengthen the main information asymmetry andreduce the uncertainty of the environment, improve the predictive ability of the forecastthrough access to information and improve the predictive ability.(1) The basic theory and models of collaborative demand forecast in supply chainare built in thesis. The order parameter, synergy, damping (resistance), as well as thestability of coordination is analyzed based on Synergetics. The evolution mechanism ofdemand information in supply chain is studied from four dimensions, such as perception,cognition, decision-making and command. The demand information evolution model andintegrated collaborative demand forecasting model with feedback feature are presented.Complexity and key success factors for collaborative demand forecasting are discussedbased on technology acceptance model.(2) Information and methods collaboration of demand forecasting in supply chainare studied. Information collaboration are divided into four types based on the charac-teristics of multi-source information and the amount of information, such as inadequateinformation, information complementing, information redundancy and information ex-clusive. A collaborative demand forecast method is proposed for demand forecastingbased on historical information and real-time information. And also the demand forecast-ing methods based on multi agent and multi methods fusion are studied in the thesis. Asupport vector machines and rough sets based improved forecasting method is presented to improve demand forecasting performance and accuracy. In order to further improvethe of the forecasts, comprehensive utilization of the environment, competitors, macro-economic conditions, etc.,are used to correct the demand forecast based on evidence andBayesian networks.(3) Dynamic demand forecasts are studied based on the demand characteristics andconsumer perspectives. The study discusses the widespread demand for the namely cycli-cal phenomenon, analyzes the effect of predict frequency for the forecast. Because thegood has the characteristics of storable and irreplaceable, the demand will increase ob-viously when the price has changed (i.e. sale activity etc.). The consumer likes to buymore goods when the price is lower and stores the goods for future consuming, which iscalled consumer inventory. The study discusses the phenomenon above and present somemodels to explain the phenomenon based on utility theory, consumer economics and sys-tem dynamics modeling. As well as the impact on the demand forecasts is discussed, anda comprehensive forecast method based on POS data is proposed to make more accurateand dynamic forecast.(4) As the materials and the information has significant delay when they transmittedin the supply chain, it needs to be agile to achieve an effective collaborative demand fore-casts. Agile demand forecast is studied from organization, partnerships and informationsystems perspectives, the advantages and disadvantages of the traditional organization insupply chain is discussed and a new team-based ?exible form of organization is presentedto support agile demand forecasting. Demand forecast under agile partnerships is ana-lyzed and a method for supply chain partners selection is designed based on AHP andNEULONET.(5) The methods and processes for collaborative demand forecasting are analyzedand designed based on multi-source information fusion and collaboration. An integratedmodel is presented, which includes six steps: data, organization, environment and excep-tions, Decision-making and methods, coordination and adjustment of the corresponding.The key processes are modeled and simulated with Petri nets. Finally, an agile frame-work model for the demand forecasting is proposed based on multi-agent and Internet tosupport collaborative demand forecasting.The study can eliminate the uncertainty of demand forecasts, improves forecast ac-curacy and optimize the allocation of resources. The constructed theory, models and methods can enrich the supply chain management theory and guide the supply chain prac-tice, it will have a positive theoretical and practical significance.