Solving supplier selection problem by fuzzy clustering and neural network techniques

Companies have increased their level of out-sourcing and are relying more heavily on their supply chain as a source for their competitive advantage. Thus, determining which suppliers to include in the supply chain has become a key strategic consideration. For example, in industrial companies, according to studies, purchasing share in the total turnover typically ranges between 50 to 90%, indicating the importance of supplier selection decisions. Many articles have addressed this problem area, but only a few encompass the complete scope of stages involved in selecting a supplier. In this thesis an entire Supplier Selection procedure is developed based on a set of assumptions. The Supplier Selection procedure is divided into three phases and the design recommendations for implementation of two of these phases are provided along with a supervised PFCM (Possibilistic fuzzy c-means) clustering algorithm. Also provided in the thesis are guidelines for relaxing the assumptions of the model using CCT (Collaborative Control Theory) principles of Collaborative e-Work parallelism, JLR (Join/Leave/Remain) and CRP (Collaboration Requirement Planning). The PFCM clustering algorithm was tested on five different datasets (three randomly generated, Iris data set, and supplier data set from published literature). The new PFCM algorithm accurately predicted 3 as the number of clusters for the randomly generated and Iris data sets. PFCM was then validated using the supplier data set and had 77.8% matching with the supplier clusters suggested by MMR (min-min roughness). This thesis presents a new supplier selection protocol based on supervised clustering algorithm which is validated with five different datasets. It also provides a collaborative insight in the supplier selection decision making process applying CCT principles