An Analysis of Retail Supply Chains; Simulation in Neural Networks and Maximum Flow Networks

Contemporary supply chain management systems need not only be adaptive to changing needs of the consumers but also to the varying terms amongst the members of the supply chain. Moreover, the on- time availability of services and goods is imperative for retaining customers. Hence, irrespective of a new supplier becoming a partner of the supply chain or an old supplier exiting the business, the demand must be met keeping cost constraints into account. Supply chain management is a cross-functional approach that involves managing the movement of raw materials into an organization, processing of materials into finished goods and finally providing finished goods to the end consumers. In this paper, the concept of SCM has been perused with the approach of network flows applying maximum flow network algorithm of graph theory with changing count of the supply chain players and their supplying capabilities. We have discussed cases of increasing count of Suppliers, Manufacturer, Distributors and Retailers from 8 to 88 players of an apparel retail chain. This variability is then analyzed for time usingMATLAB Simulation tool. The same variability is assessed using feed forward neural networks for modifying the number of neurons within the various layers.This study emphasize on the analysis of the performance of the two approaches which calculate the material flows across several alternate supply sub-chains from suppliers to the destined retail stores. The algorithms are simulated with varying data to compute actual supply chain logistics and then, the running time of the algorithm is compared in a dynamic node scenario. The results of the two approaches have been compared using mean squared error of the difference between actual demand and expected demand as well as running time of these two approaches. Simulation of supply chains has been performed using weighted graphs with non- negative edges and employing iterative Ford Fulkerson algorithm and multilayer feed forward network using inbuilt MATLAB functions