This research has led to a novel methodology for assessment and quantification of supply risks in the supply chain. The research has built on advanced Knowledge Discovery techniques and has resulted to a software implementation to be able to do so. The methodology developed and presented here resembles the well-known consumer credit scoring methods as it leads to a similar metric, or score, for assessing a supplier’s reliability and risk of conducting business with that supplier. However, the focus is on a wide range of operational metrics rather than just financial, which credit scoring techniques typically focus on. The core of the methodology comprises the application of Knowledge Discovery techniques to extract the likelihood of possible risks from within a range of available datasets. In combination with cross-impact analysis, those datasets are examined for establish the inter-relationships and mutual connections among several factors that are likely contribute to risks associated with particular suppliers. This approach is called conjugation analysis. The resulting parameters become the inputs into a logistic regression which leads to a risk scoring model the outcome of the process is the standardized risk score which is analogous to the well-known consumer risk scoring model, better known as FICO score. The proposed methodology has been applied to an Air Conditioning manufacturing company. Two models have been developed. The first identifies the supply risks based on the data about purchase orders and selected risk factors. With this model the likelihoods of delivery failures, quality failures and cost failures are obtained. The second model built on the first one but also used the actual data about the performance of supplier to identify risks of conducting business with particular suppliers. Its target was to provide quantitative measures of an individual supplier’s risk level. The supplier risk scoring model is tested on the data acquired from the company for its performance analysis. The supplier risk scoring model achieved 86.2% accuracy, while the area under curve (AUC) was 0.863. The AUC curve is much higher than required model’s validity threshold value of 0.5. It represents developed model’s validity and reliability for future data. The numerical studies conducted with real-life datasets have demonstrated the effectiveness of the proposed methodology and system as well as its future potential for industrial adoption.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:675883 |
| Date | January 2015 |
| Creators | Chuddher, Bilal Akbar |
| Contributors | Makatsoris, H. |
| Publisher | Brunel University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://bura.brunel.ac.uk/handle/2438/11628 |
Page generated in 0.0027 seconds