Applying two-stage SOM-based clustering approaches to industrial data analysis

Abstract

Data analysis is a promising method for reducing the complexity of information management when handling huge amounts of data. In this paper the performances of hybrid two-stage methods combining self-organising map (SOM) and traditional clustering algorithms are presented and evaluated with the goal of identifying the techniques leading to the best clustering quality. The SOM-based two-stage methods are also compared to single-stage approaches applying traditional hierarchical and partitioning algorithms. These comparisons are initially based on the analysis of two reference data sets (Iris and Abalone) which shows how the use of SOM improves clustering quality while reducing computational time. In order to further evaluate the proposed two-stage method, the comparison is extended to two industrial applications. The first one concerns the group technology problem and the second is related to the classification of purchased components. The obtained results show that SOM + K-means can achieve a satisfying clustering assignment quality while reducing the computational time.

Keywords:

• Data analysis
• Clustering
• Self-organising map
• Group technology
• Product classification

Acknowledgements

The authors thank the Swiss Agency for Promotion and Innovation for the financial support of this work.

Luca Canetta is a PhD student at the Laboratory for Production Management and Processes at the Swiss Federal Institute of Technology at Lausanne. In 1999 he received his engineer diploma in Management and Production Engineering at the Department of Industrial Engineering of the University of Bergamo, Italy. His research interests include customer behaviour analysis, neural networks applications and electronic commerce.

Naoufel Cheikhrouhou is a senior scientist in Operations Management and Simulation at the Swiss Federal Institute of Technology at Lausanne, where he is leading a group on Operations Management. He received his industrial engineer diploma from the Ecole Nationale d’Ingenieurs de Tunis, Tunisia and his PhD in Industrial Engineering from the Institut National Polytechnique de Grenoble, France. His main research interests are in the area of modelling, simulation and optimisation of production networks, the integration of human factors in production management and the reduction of management complexity in logistics and services. Due to his work on these aspects, Dr. Cheikhrouhou received the Burbidge award in 2003. Leading different projects with the collaboration of industrial partners, Dr. Cheikhrouhou has also published several papers in various journals and international conferences.

Remy Glardon is full professor at the Swiss Federal Institute of Technology at Lausanne where he teaches Production Planning and Control, Materials Selection and Production Processes. He also teaches the Production Systems Module of the Postgraduate MS-Program in Management of Logistical Systems and the Industrialisation and Production Module of the Postgraduate MS-Program in Management of Technology. He was the director of the former Institute for Design, Analysis and Production of Mechanical Systems. He is the director of the Laboratory for Production Management and Processes. He received his MS in Mechanical Engineering in 1973 and his PhD in Materials Science in 1977 from the Swiss Federal Institute of Technology at Lausanne. From 1979 to 1982 he carried out postdoctoral research at UC Berkley in the field of mechanical behaviour of materials and wear mechanisms. From 1982 to 1995 he worked in various management positions in industry: R&D, quality assurance, production management, logistic and operations. His research interests are related to production planning and control, production management and logistics, production processes, rapid manufacturing and selective laser sintering. He is a member of ASME, ASM and SVMT (member of the steering committee 1994 to 1998). He is a co-founder and a member of the steering committee of the Human Technology Organisation (HTO) Swiss network. Professor Glardon has also published several papers in various journals and international conferences.