Knowledge discovery from supplier change control data for purchasing management

The self-organizing map (SOM) is a powerful neural network tool for analyzing multivariable data. It reveals the interrelations within the variables through an iterative learning process. The clustering and topology preserving properties have made the SOM an ideal tool to exploring large datasets (large in both attributes and data records). This paper focuses on using a real life manufacturing dataset about changes to a product or process made by suppliers of the company. We use results from this analysis to show what SOM can provide as in depth understanding of the dataset. We also provide techniques to encode symbolic variables into forms that the SOM can admit. Procedures are also provided to interpret the SOM output results