Improved Pattern Discovery in Supply Chain Management with Graphic Processing Unit (GPU)

Improved Pattern Discovery in Supply Chain Management with Graphic Processing Unit (GPU)

Lau Mai Chan^a, Rajagopalan Srinivasan^a,b

^a Department of Chemical and Biomolecular Engineering, National University of Singapore, Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore

^bProcess Sciences and Modeling, Institute of Chemical and Engineering Sciences, Pesek Road, Jurong Island, Singapore 627833, Singapore

Abstract

While the use of supply chain management (SCM) is getting growing attention from all range of businesses, researchers are continue expanding the scope of supply chain management in numerous dimensions so as to achieve further improvement in organizational performance as well as to sustain the competitive position. Advanced supply chain management system attempts to embrace as many factors which are having potential impact on business decisions from various aspects, including operational, purchasing and delivery. The scope expansion is also going beyond the firm as the hierarchy of SCM has been observed growing externally which include indirect entities, for instance indirect suppliers which are one level higher than the direct suppliers. The adoption of supply chain management system entails business entities the competitive advantage only if meaningful and relevant knowledge are captured, which in turn enables reliable prediction to be made on business. However, the expansion of SCM scope has led to the generation of huge amount of data which makes the knowledge discovery process difficult. Owing to the limitation of computing resources, the conventional sequential data processing techniques are inefficient if not unable to handle the gigantic amount of data. As a result, parallel computing technologies evolve to be a potential solution for big data analysis. Among all the parallel computing resources such as supercomputers, grids and cluster, our group is particularly interested in harnessing the computing power from General Purpose Graphic Processing Unit (GPGPU) because of the relatively low cost and availability as commodity. On top of that, Graphic Processing Units consist of a number of Single Instruction Multiple Data (SIMD) multi-processors which map well with the implementation of many data analysis and pattern discovery methods. The aim of this work is to develop GPU parallel algorithm(s) which is (are) able to capture higher quality knowledge for business decision making purpose by having higher degree of efficiency. The quality and efficiency of the parallel algorithm will be examined with a simulated SCM system.

Keywords: Graphics Processing Unit (GPU) parallel computing, Supply Chain Management (SCM), Pattern Discovery